[Frontiers in Bioscience 3, d1148-1160, November 15, 1998] |
MOLECULAR PATHOBIOLOGY OF PANCREATIC ADENOCARCINOMA
Shamlal Mangray and Thomas C King
Department of Pathology and Laboratory Medicine, Brown University School of Medicine, Providence, RI
Received 8/24/98 Accepted 9/4/98
5. GENETIC ALTERATIONS IN PANCREATIC ADENOCARCINOMA
The most commonly altered genes in pancreatic adenocarcinoma are: K-ras (74-100%), p16INK4a (up to 98%), p53 (43 to 76%) and DPC4 (53%) (20-24). As shown in table 2 K-ras is a proto-oncogene whereas all the others are tumor suppressor genes. Additional genes have been found to be altered at lower frequency and some as yet uncharacterized genetic lesions may be important in tumorigenesis. Understanding the technical aspects of genetic analysis in these tumors is important in interpreting the significance of reported data and in designing rational diagnostic and therapeutic strategies. Assessment of proto-oncogene mutations such as K-ras is straight forward and reliable. Since only a limited number of genetic changes are expected, these can usually be assessed with good sensitivity. The same is not true for most tumor suppressor gene alterations which may correspond to homozygous deletions or novel point mutations. Methodologies for screening for p53 gene mutations are well established and have good sensitivity both for identifying novel mutations and for detecting mutations in a relatively small percentage of cells in a tissue sample. Unfortunately, detection of homozygous deletion (a common alteration for p16INK4a) or promoter mutations (which occur in Rb1) is much more problematic. In particular, deletion analysis is hampered by the presence of normal tissue in biopsy specimens (endothelial cells, fibroblasts) which contribute to background and can lead to false negative results.
Table 2. Oncogenes and Tumor Suppressor Genes Commonly Altered in Pancreatic Adenocarcinoma
* Calculated from reference 21 assuming that reported p16INK4a gene deletions and point mutations but not promotor hypermethylation result in loss of p19ARF. ** Calculated from reference 21 assuming that reported p16INK4a gene deletions but not point mutations and promotor hypermethylation result in loss of p15INK4b.
For deletion analysis of tumor tissue, it is necessary to quantitatively assess the amount of DNA present in the sample to determine whether homozygous deletion has occurred. Since tumor necrosis and formalin fixation damage DNA to a variable extent in different tumor biopsies, it is necessary to monitor for DNA quality as well. This can be accomplished by employing a synthetic competitive substrate lacking a Kpn I restriction enzyme site present in the native p16INK4a sequence. A constant amount of this competitor is added to PCR reactions containing known amounts of tumor DNA. PCR products made from the two templates can then be distinguished by digestion with Kpn I followed by gel electrophoresis as shown in figure 1A. The ratio of the intensity of the two bands reflects the number of intact copies of the gene present in the sample. This ratio is insensitive to PCR conditions and allows a true estimate of the number of amplifiable gene copies present in a sample. To control for the quality of DNA in different samples, a second competitive PCR assay is performed using identical amounts of sample DNA with primers for a control gene which produce a PCR product of nearly identical size. A decrease in normalized p16INK4a gene copy number of more that 50% is then indicative of homozygous deletion in the tumor sample as shown in figure 1B.
Figure 1A. Autoradiograph of gel electrophoresis of p16INK4a competitive PCR assay after restriction digestion. A constant amount of competitor was added to each reaction with the indicated amount of control genomic DNA added. Note equivalence of band intensities with 1.0 nanogram of control DNA.
Figure 1B. Autoradiograph of gel electrophoresis of p16INK4a and control (Factor V) competitive PCR assays after restriction digestion. Two tumor samples, 1 & 2, are shown in each PCR assay. Tumor 1 shows no evidence of p16INK4a deletion with equal band intensity ratios for the p16INK4a and control (Factor V) PCR assays. Tumor 2 shows evidence of homozygous p16INK4a gene deletions with complete loss of the signal for genomic p16INK4a with preservation of the control (Factor V) genomic band.
The use of tissue culture cell lines or xenografts from tumors can overcome the difficulties posed by contaminating normal tissue by producing a pure tumor cell population. Unfortunately, these techniques have been shown to select for in vitro deletion of the gene of interest in some tumors including pancreatic adenocarcinoma and can thereby overestimate the frequency of gene deletions (25). These alternate methodologies clearly have the potential to significantly under or over estimate the frequency of homozygous deletion and these technical limitations must be considered when evaluating the literature.
5.1. K-ras proto-oncogene
The K-ras gene which is located on chromosome 12p13 has been found to be mutated in many precursor lesions associated with pancreatic adenocarcinoma as well as in the vast majority of pancreatic adenocarcinomas. Mutations have been identified in up to 28-81% of dysplasias and noninvasive intraductal lesions (4,5). The differences in frequency of K-ras mutations may be explained by the inclusion of low and high grade dysplasia and adenomas in different studies. These findings are consistent with a model in which K-ras gene mutations are an early and essential step in the development of most pancreatic adenocarcinomas.
The K-ras gene encodes a guanine nucleotide binding protein involved in signal transduction. Abnormal activation of the K-ras oncogene can occur as a result of point mutations at codons 12,13 or 61. Specific amino acid substitutions at these sites then alter it's GDP/GTP binding site and result in retention of GTP (by blocking it hydrolysis to GDP) producing a constituitively active K-ras protein. One of the steps in the normal attachment of K-ras to the cell membrane involves farnesylation of the ras protein by farnesyl transferase. Without this step, the ras protein cannot associate with the cell membrane and so cannot mediate signal transduction. Blockage of farnesylation can then potentially inhibit both normal signal transduction as well as cell transformation by mutant ras proteins. This pathway may have implications for treatment as discussed below.
Hruban et al. observed point mutations in 68 of 82 (83%) of surgically resected or biopsied adenocarcinomas of the pancreas using primer mediated, mutant-enriched, polymerase chain reaction-restriction fragment length polymorphism analysis and allele-specific oligonucleotide hybridization (26). 49% of these mutations were guanine to adenine transitions, 39% guanine to thymine transversions, and 12% guanine to cytosine transversions. Of note, the overall prevalence of K-ras mutations in patients with a history of smoking at some point in their lives was 88% versus 68% in patients who never smoked.
5.2. p16INK4a /p15INK4b/p19ARF tumor suppressor genes
The INK4 family of cyclin dependent kinase (CDK) inhibitors is comprised of multiple gene products which repress cell cycle progression at G1S (27-29) (figure 2). Loss of these gene products then favors cell cycle progression at G1S. p16INK4a (MTS-1, INK4A) was the first member of the INK4 family identified and it is altered in a significant percentage of human tumors (30). The p16INK4a gene product inhibits the interaction of cyclin D with CDK4. This cyclin D-CDK4 complex phosphorylates the retinoblastoma gene product (Rb1) preventing the formation of the E2F-Rb1 complex (figure 2). This phosphorylation step frees E2F to act as a transcription factor with the resultant progression of the cell cycle to S phase. Loss of p16INK4a activity results in the elimination of these inhibitory effects at the level of cyclin D-CDK4 interaction, thereby promoting cell cycle progression (31). p16INK4a deletion is associated with progression in follicular center cell lymphoma to high grade lymphoma and with decreased survival in nonsmall cell lung cancer (32,33).
Figure 2. Cells cycle control pathways altered in pancreatic adenocarcinoma. Schematic diagram of gene products controlling entry into the cells cycle (G0 --> G1) and G1S progression. Genes frequently altered in pancreatic adenocarcinoma are in magenta with their chromosomal location noted. Normal positive and negative regulatory interactions are indicated in green and red respectively. Gene mutation typically results in the loss of a negative regulatory signal (e.g. p16INK4a) or the enhancement of a positive signal (e.g. K-ras).
Loss or decreased activity of p16INK4a can occur through a variety of mechanisms. Many tumors with p16INK4a alteration show evidence of homozygous deletion at this locus. In some tumors point mutations in the p16INK4a coding sequence are more frequent than deletions while in other tumor types, inactivation of the p16INK4a promotor by methylation has been reported without evidence of mutation or deletion of the structural gene (34).
The p16INK4a gene located on chromosome 9p21 is composed of three exons: a 5I region of 126 bp (coding exon 1), a middle region of 307 bp (coding exon 2) a 3 region of 11 bp (coding exon 3). A second INK4 family member is closely linked to p16INK4a. The p15INK4b gene is located 30 kb downstream from the p16INK4a gene and is deleted in most tumors with p16INK4a deletions (35). p15INK4b is upregulated by TGF beta (Fig 2) and inhibits the formation of activated CDK4 in the same manner as p16INK4a (36). Another putative tumor suppressor gene is imbedded in the p16INK4a locus. p19ARF derives in part from an alternate reading frame in the p16INK4a gene. It utilizes the same second exon as p16INK4a with a separate first exon and promotor. p19ARF is therefore inactivated in all tumors with p16INK4a gene deletions and in many but not all tumors with p16INK4a point mutations (37). The biologic significance of the unusual genetic organization of the p16INK4a locus is unclear. Recently, p19ARF has been shown to bind to the mdm2 protein and thereby inhibit its interaction with p53 (38,39). Normally, mdm2 regulates p53 by binding to it and accelerating its degradation and blocking its effects on gene transcription. Overexpression and /or gene amplification of mdm2 in some types of tumors (but not in pancreatic adenocarcinoma) results in functional loss of normal p53 function and a transformed phenotype. Loss of p19ARF then mimics the effects of mdm2 overexpression by permitting mdm2 to freely interact with p53 and down regulating it. Consequently, most homozygous deletions at this locus result in the coordinate loss of three gene products involved in regulating the G1S checkpoint with potentially synergistic facilitation of cell cycle progression. No other single genetic target has the potential to result in this number of genetic alterations targeting and single cellular pathway.
p16INK4a gene deletions and/or mutations have been found in a significant number of pancreatic cancers but their prevalence varies substantially in different reports (27-82%) (40). Schutte et al. suggested that p16INK4a gene alterations are present in virtually all pancreatic carcinomas when promotor inactivation by methylation is considered as well as gene mutation and deletion (41). These investigators found alterations in p16INK4a in 49 of 50 pancreatic carcinoma xenografts and cell lines.
This viewpoint is controversial, however, since other investigators have found substantially lower frequencies of p16INK4a gene alterations (27% and 38%) in primary pancreatic adenocarcinomas (22,42). Huang et al. carried out deletion and mutational analyses of p16INK4a in 30 microdissected primary human ductal pancreatic carcinomas from patients who were not subjected to radiotherapy or chemotherapy prior to surgical resection of the tumors, and 18 human pancreatic carcinoma cell lines (22). Single-strand conformation polymorphism and DNA sequence analyses, and PCR-based deletion analyses were performed for mutations and homozygous deletions of the p16INK4a gene respectively. The overall frequency of deletions and mutations was 66% for the cell lines compared to 27% for primary tumors. It is clear from the study of other types of human tumors that the frequency of detectable p16INK4a gene alterations is substantially increased in tumor cell lines as an artifact of in vitro culture (43). Observed differences in the frequency of p16INK4a deletion by different investigators is likely methodologic (see above) and that the true frequency of p16INK4a gene deletion in primary pancreatic cancers is uncertain (25).
In addition to gene inactivation by intragenic mutation and gene deletion, transcriptional inactivation of the p16INK4a gene secondary to hypermethylation of its 5'-CpG island was investigated by methylation-specific PCR. In one study using xenografts and cell lines, seven of nine pancreatic adenocarcinomas without demonstrable p16INK4a mutations showed evidence of hypermethylation. Conversely, nine tumors with identified p16INK4a mutations did not show evidence of hypermethylation (41).
5.3. p53 tumor suppressor gene
The human p53 gene, located on chromosome 17p13, is a tumor suppressor gene that plays a central role in genetic stability and cell survival. p53 gene mutations are common in pancreatic adenocarcinoma (43-76%). Germline p53 gene mutations (Li-Fraumeni syndrome) result in early onset tumors of various types and although the risk of pancreatic adenocarcinoma is low compared to the incidence of other malignancies, it is still substantially greater than in the general population. Li et al identified 4 cases of pancreatic cancer in 151 patients with Li-Fraumeni syndrome (2.6%) (44). This cohort of patients had a total of 169 primary tumors. Interestingly, three of the four pancreatic tumors occurred in individual over 45 years of age, similar to the demographics of sporadic pancreatic adenocarcinoma. The shortened life expectancy of these patients may have contributed to the relatively low incidence of pancreatic adenocarcinoma in these patients.
The p53 tumor suppressor gene serves a critical role at the G1S transition by blocking cell entry into S phase in response DNA damage. The p53 gene has been proposed to exert an inhibitory effect on CDK4 directly, and indirectly through p21/WAF1 as shown in figure 2. Wild-type p53 is also necessary for the efficient activation of apoptosis in sensitive cells in response to DNA damage (45). Apoptosis is the major mechanism by which ionizing radiation and many chemotherapeutic agents cause tumor cell death. There is now substantial in vitro and in vivo evidence that p53 mutations are associated with decreased response to many types of chemotherapy and may be a predictor of decreased survival in some tumors (46-49). This has been supported by studies in pancreatic carcinoma cell lines whereby reintroduction of wild-type p53 activates apoptosis (50). Treatment protocols which aim to restore normal p53 function in tumor cells by gene therapy are now underway in some centers. p53 mutations may also result in phenotypic changes in tumor cells by virtue of its role as a transcription factor which may affect the aggressiveness of pancreatic carcinoma (50).
Numerous different p53 point mutations can occur and their precise phenotype cannot always be predicted from their DNA sequence. Mutations in the p53 gene typically result in an abnormal protein that is longer-lived than the wild-type protein. Because p53 normally functions as a dimer, mutant p53 protein can sequester wild-type protein in nonfunctional heterodimers resulting in a dominant negative phenotype for many p53 point mutations (51). Many tumors with p53 mutations develop loss of heterozygosity (LOH), discarding the wild-type p53 allele. The long half-life of mutant p53 has allowed its detection by antibody staining of tissue sections (52). It has become clear, however, that overexpression of wild-type p53 can occur in some tumors so that immunochemical detection of p53 is not necessarily indicative of p53 gene mutation (53).
5.4. DPC4 (Smad4) tumor suppressor gene
DPC4 is located on chromosome 18q21 and shows homology to the mad family of proteins which play a critical role in signal transduction through the TGF beta superfamily of cell surface receptors. They are thought to act through p15INK4b to inhibit CDK4 interaction with cyclin D. Therefore inactivation of this gene promotes progression from the G1S restriction point as described above (figure 2) . Recent evidence suggests that TGF beta expression may be a positive prognostic factor in pancreatic adenocarcinoma (54).
This gene shows biallelic inactivation in approximately 50% of cases of pancreatic adenocarcinoma. In most cases, DPC4 is inactivated by homozygous deletion with point mutation followed by loss of heterozygosity in the remaining cases. Rozenblum et al noted a high concordance for DPC4 and p16INK4a inactivation. Interestingly, all of these cases with DPC4 inactivation also had p16INK4a inactivation but not vise versa (21). These findings are consistent with a model in which genetic inactivation of p16INK4a increases the selective advantage of developing a DPC4 mutation. Since DPC4 appears to act primarily through p15INK4b, it would be surprising if DPC4 deletion would be effective if p15INK4b was deleted with p16INK4a (figure 2) .
5.5. FHIT, fragile histidine triad, candidate tumor suppressor gene
The FHIT gene encodes an Ap3A hydrolase, a member of the histidine triad gene family. Diadenosine oligophosphates appear to act as cytokines and their concentration is normally elevated in cells at G1S (55). The FHIT gene product can inactivate the signaling function of diadenosine oligophosphates by hydrolyzing them. Inactivation of FHIT may then promote cell cycle progression at G1S by virtue of higher diadenosine oligophosphate levels. FHIT has been localized to the 3p14.2FRA3B site which is targeted for homozygous deletion in colon and gastric carcinoma and pancreatic adenocarcinoma (56). Ohta et al identified aberrant FHIT transcripts in these tumors, suggesting a role for this gene in carcinogenesis (57). Various investigators have implicated FHIT as a putative tumor suppressor gene in malignancies such as breast, renal and lung carcinoma and a similar role in pancreatic adenocarcinoma has been suggested. Simon et al found that full-length FHIT transcripts were absent in 10 of 14 pancreatic carcinoma cell lines (71%). Most of these cases had intragenic homozygous deletions of FHIT exons 3, 4 and 5 (23).
Some have argued that FHIT alterations are not functionally important in tumorigenesis but simply result from its coincident location at an active fragile site (58). This innocent bystander hypothesis is plausible since most of the reported data derives from human tumor cells lines. Aberrant FHIT transcripts observed in such cell lines may be an in vitro artifact in some cases. Indeed, aberrant FHIT transcripts have been observed in non-tumor tissues and are more common in aging cells. Some of these abnormalities may be due to reduced RNA-splicing fidelity in aging cells.
Conversely, the frequent loss of heterozygosity at 3p14 in premalignant conditions argues for a role for FHIT in tumorigenesis (59). Experiments transfecting wild-type FHIT into tumor cell lines with FHIT abnormalities have produced conflicting results. Otterson et al found that stable overexpression of wild-type FHIT did not alter cell morphology, inhibit colony formation, or inhibit cell proliferation in vitro (60). In addition overexpression of wild-type FHIT did not lead to altered cell cycle kinetics in dividing cells. In contrast, Siprashvili et al found that transfection of wild-type FHIT into several tumor cell lines resulted in a reduction of the frequency and size of tumors developing after transfer into nude mice (61). These divergent results can be reconciled if the effects of FHIT were only exposed in vivo, in nude mice, and were not evident in tissue culture. Clearly more information is needed to decide this question and whether there could be a role for FHIT in gene therapy. The effects of FHIT replacement in pancreatic adenocarcinoma have not been reported.
5.6. HER-2/neu (erbB2) proto-oncogene
HER-2/neu is a proto-oncogene that encodes a transmembrane glycoprotein with tyrosine kinase activity, which is closely related to the epidermal growth factor (EGF) receptor. Up regulation or amplification of this gene has been associated with a poor prognosis of cancers of the breast and ovary. Increased expression of HER-2/neu has been observed in noninvasive intraductal lesions of the pancreas as well as pancreatic adenocarcinoma.
Yamanaka et al showed that both HER-2/neu protein and mRNA are detectably expressed in the acinar cells and ductal cells of normal human pancreas. HER-2/neu immunostaining was demonstrated in normal pancreas and in 34 of 76 (45%) pancreatic carcinomas. They also found that HER-2/neu mRNA was overexpressed in pancreatic adenocarcinoma in the absence of HER-2/neu gene amplification assessed by Southern hybridization (62).
Other investigators did not find consistent immunohistochemical evidence of HER-2/neu expression in normal pancreatic tissue but did find expression in some tumors and intraductal lesions. HER-2/neu expression was demonstrated in 82% of ducts with flat mucinous hyperplasia, 86% of ducts with papillary mucinous hyperplasia without atypia, 92% of ducts with atypical papillary mucinous hyperplasia and all specimens with carcinoma in situ (63). HER-2/neu expression was absent in all but one case of normal pancreatic ducts and ductules. HER-2/neu expression was observed in 69% of moderately differentiated infiltrating carcinomas but none of the poorly differentiated carcinomas.
Dugan et al also showed increased expression of HER-2/neu in pancreatic adenocarcinoma by immunohistochemical methods (64). In a series of 79 primary pancreatic adenocarcinomas, they found significantly lower expression of HER-2/neu in poorly differentiated areas of tumor compared to well and moderately differentiated areas. Most well and moderately differentiated invasive tumors (62%) showed moderate or strong expression of HER-2/neu while a minority of poorly differentiated tumors (19%) showed similar expression (63).
The observed differences in staining of normal pancreatic parenchyma almost certainly result form differences in the sensitivity of the immunohistochemical stains employed. Antibodies for HER-2/neu have been problematic in the evaluation of breast carcinoma because of the effects of tissue fixation on antigen retrieval and the lack of internal controls for staining. It will be of interest to see if the reported lack of HER-2/neu gene amplification in pancreatic adenocarcinoma is confirmed by FISH. Available evidence suggests that HER-2/neu overexpression is a relatively early event in the development of pancreatic adenocarcinoma which is lost in advanced, poorly differentiated tumors. If this model is validated in subsequent studies, biologic therapy with HerceptinTM (human anti-HER-2/neu antibody) may not be of much advantage in patients with poorly differentiated pancreatic adenocarcinoma.
5.7. RCA2, hereditary breast Cancer
BRCA2 has been implicated as a potential tumor suppressor in familial aggregates of pancreatic carcinoma with a history of breast carcinoma. Goggins et al. have also shown germline BRCA2 gene mutations in patients with apparently sporadic pancreatic carcinomas(65). In an unselected panel of 41 pancreatic carcinomas (30 xenografts and 11 pancreatic cell lines) screened for alterations of BRCA2, 15 (27%) showed loss of heterozygosity. Four (10%) had abnormalities in the second allele when the entire BRCA2 gene was screened using a protein truncation assay. Three of these four mutations were identified as germline in origin. One of the patients with a germline mutation had a single relative with breast cancer while another had a single relative with prostate cancer. None had a family history of pancreatic adenocarcinoma. Based on these limited data, the incidence of germline mutations of BRCA2 in pancreatic adenocarcinoma may be as high as in breast and ovarian cancer. The significance of loss of heterozygosity for BRCA2 in some cases of pancreatic adenocarcinoma is less clear since loss of heterozygosity has not been found in breast cancers from BRCA2 kindreds.
5.8. Rb1, retinoblastoma tumor suppressor gene
Rb1 has been shown to be altered in less than 10% of pancreatic carcinomas investigated (66,67). Due to the large size of the gene and the nature of alterations in it in different tumors, it may be difficult to reliably identify mutations and this is probably a minimum estimate for the number of Rb alterations in pancreatic adenocarcinoma. Tumors with p16INK4a alterations do not usually have coexistent Rb mutations since these would tend to mitigate the effects of p16INK4a loss. Since p16INK4a alterations are common in pancreatic adenocarcinoma, Rb mutations are expected to be uncommon (41)