Context.—Metastatic disease is the most critical determinant of resectability of pancreatic cancer and accounts for the poor outcome of patients with this disease. Thus, a better understanding of metastatic pancreatic cancer will afford new opportunities for therapeutic intervention.
Objective.—To summarize and discuss the current understanding of the clinical and molecular features of metastatic pancreatic cancer.
Data Sources.—Published literature on advanced stage pancreatic cancer, pancreatic cancer metastasis, and autopsy findings in patients with pancreatic cancer.
Conclusions.—In the clinical setting, it can be difficult to distinguish a metastatic pancreatic carcinoma from primary neoplasms in the liver, lung, or ovary. However, immunolabeling for DPC4 protein as part of a diagnostic panel is useful for making this distinction. Emerging data from a variety of investigators now indicate that overexpression of EphA2, loss of DPC4 and MKK4, and aberrant activation of the Hedgehog signaling pathway are associated with metastatic propensity of pancreatic cancers, providing novel therapeutic targets for the most lethal stage of this disease.
Pancreatic cancer is the fourth leading cause of cancer-related mortalities in the United States. In the year 2008 in the United States, an estimated 37 680 patients will be diagnosed with pancreatic cancer, and 34 290 patients will die of their disease.1 These unfortunate statistics reflect the advanced stage at which most patients with pancreatic cancer are diagnosed and the paucity of effective chemotherapeutic regimens for advanced disease.2–4 Thus, a better understanding of the factors that contribute to pancreatic cancer progression and metastasis is critical in combating this disease.
Contrary to early pancreatic carcinogenesis,5 the pathology and molecular features of advanced stage pancreatic cancers are relatively unexplored. As a result, little is known about the mechanisms responsible for uncontrolled local growth and metastatic spread, the very processes ultimately responsible for most pancreatic cancer– related deaths.6 This is underscored by the fact that unlike advanced stage colorectal cancers in which the primary cancer and/or liver metastases are commonly resected for clinical benefit and are thus available for medical research, the management of advanced stage pancreatic cancer often does not include surgical resection.7,8
A better understanding of the clinical, genetic, and molecular features of metastatic pancreatic cancer will afford novel areas for therapeutic intervention. Toward this goal, we provide a review of the literature on the pathology and genetics of advanced stage pancreatic cancer including recent advances that may lead to novel treatments for this most lethal stage of the disease.
Distant metastases are a frequent occurrence in patients with pancreatic cancer.9–12 The most common site of distant metastasis found at autopsy is the liver, followed by the peritoneum, lung and pleura, bones, and adrenal glands (Figure 1).9–12 However, metastatic pancreatic cancer has been reported in virtually every organ site, including the brain, diaphragm, gallbladder, heart, small and large intestines, kidneys, ovaries, pericardium, seminal vesicles, skin, stomach, spleen, testis, thyroid gland, urinary bladder, and uterus.9,10,12 Calculations based on estimates of tumor doubling times indicate that occult liver metastases are often present at the time of pancreatectomy.13 Even small (<2 cm) pancreatic cancers are associated with metastases,10 suggesting metastatic propensity is rapidly acquired during pancreatic carcinogenesis and progression. Consistent with this notion, molecular analyses indicate micrometastases are frequently present within the peritoneal cavity and bone marrow of patients with potentially resectable pancreatic cancer.14–16
Although the liver is the most common site of distant metastasis, up to one-third of patients do not have hepatic metastases at autopsy. In these patients, the lungs and abdominal lymph nodes are the most common sites of metastasis possibly due to hepatofugal portosystemic shunting (ie, flow directed away from the liver to the systemic circulation) induced by splenic vein obstruction or to the presence of aggressive features of the primary carcinoma.9–11 Mao et al10 found that carcinomas of the body and/ or tail of the pancreas are more frequently associated with hematogenous dissemination than are carcinomas of the head of the pancreas. Independent of the variability among patients with respect to the sites of metastatic spread, the extent of metastatic disease at autopsy is also widely variable, ranging from less than 10 documented metastases to more than 1000 metastases.9–11,17
Hishinuma et al18 analyzed recurrence patterns for 25 autopsied patients who had undergone potentially curative resection of pancreatic cancer. Three-fourths of the patients had a local recurrence, half had hepatic metastasis, and one-third developed peritoneal dissemination. Overall, distant metastasis at any site was recognized in three-fourths of the patients. In our experience, 73% of patients who initially presented with stage I/II disease and underwent surgical resection of their infiltrating carcinomas had gross evidence of local recurrence, and two-thirds of these patients had metastatic disease.19 Although local recurrence frequently occurs following resection of pancreatic cancer,20,21 its direct relationship to patient survival is unclear.22,23
Kamisawa et al9 investigated the histologic features of pancreatic cancers in 130 patients with and without associated metastatic disease at autopsy and found that the primary cancer in patients with distant metastases tended to be more poorly differentiated. Moreover, we and others have found that advanced stage pancreatic cancers more commonly have high-grade (anaplastic) features than those reported for early stage disease.3,9,19 Of note, the finding of high-grade features at autopsy is not correlated with clinical stage or grade at initial diagnosis, indicating loss of tumor differentiation occurs with disease recurrence and/or progression.19 Winter et al24 found that loss of E-cadherin expression in pancreatic cancer is almost universally associated with anaplastic features (Figure 2), but this was not due to a methylation of the E-cadherin promoter with disease progression. In vitro models suggest that loss of differentiation in pancreatic cancer, that is epithelial-mesenchymal transition, is mediated by transforming growth factor β (TGF-β) signaling through the phosphoinositide 3-kinase/phosphatase and tensin homolog pathway.25–27
Terayama et al28 investigated the pathology of liver metastases derived from a variety of primary sites and found that small pancreatic cancer metastases frequently show an intraparenchymal, expansive pattern of growth compared with other tumor types. However, irrespective of the primary carcinoma it was derived from, the proportion of metastases showing an expansive growth pattern increased in relation to the size of the metastatic deposit. A fibrous capsule around liver metastases, which is commonly observed in liver metastases from colorectal cancer,29 was not seen in any liver metastases derived from pancreatic cancers.
In the clinical setting, pancreatic cancer metastases can pose a diagnostic dilemma in a variety of organ sites. In the liver, small metastases need to be distinguished from benign epithelial lesions, including bile duct adenomas and bile duct hamartomas (Figure 3). Bile duct adenomas are small (<1 cm), subcapsular, circumscribed, noncystic solitary lesions composed of uniform cuboidal cells that form tubules.30,31 They are believed to be a reactive process to a focal injury.30 By contrast, bile duct hamartomas (von Meyenburg complexes) are usually multiple, are cystic, contain bile, and are lined by flattened uniform “biliary” type epithelial cells.32 In the event that routine histopathologic analysis is unable to distinguish between these benign entities and a metastasis from a pancreatic cancer, immunohistochemical labeling for a variety of markers may be of value. For example, Hornick et al33 found that abnormal p53 labeling; positive labeling for B72.3/TAG-72, monoclonal carcinoembryonic antigen, and mesothelin; and loss of DPC4 labeling can help distinguish metastatic pancreatic adenocarcinoma in the liver from bile duct adenomas or hamartomas. They also found that p63 and α-methylacyl coenzyme A racemase were specific for pancreatic adenocarcinoma, but the low sensitivity of these markers may limit their value in the clinical setting. Molecular analyses such as KRAS gene status may help in more difficult cases. For example, KRAS gene mutations are present in more than 90% of metastatic cancers but occur in less than 7% of benign bile duct proliferations.34 Perhaps more problematic is the distinction between a metastatic pancreatic cancer and an intrahepatic cholangiocarcinoma. Intrahepatic cholangiocarcinoma is most commonly seen as a well- to moderately differentiated adenocarcinoma forming small tubular glands in association with a prominent desmoplastic response, an appearance not unlike pancreatic cancer (Figure 4).35 However, Argani et al36 found that intrahepatic cholangiocarcinomas uncommonly show p53 and DPC4 loss by immunolabeling compared with distal common bile duct and pancreatic cancers, providing an avenue to aid correct diagnosis of these lesions.
Metastatic pancreatic adenocarcinoma to the lung can be difficult to distinguish from primary lung cancer, a situation compounded by the fact that both tumor types have a known association with smoking.37,38 Karanjawala et al39 reported that metastatic pancreatic adenocarcinomas in the lung commonly grow along the alveoli (lepidic growth) and display a mucinous-type epithelium, thus mimicking primary mucinous bronchioloalveolar carcinoma (Figure 5). Of interest, in 20% of cases no evidence of a desmoplastic response was found even though prominent desmoplasia was a feature of the primary carcinoma and/or metastatic deposits in other organ sites. Negative immunohistochemical labeling for thyroid transcription factor 1 in association with loss of DPC4 labeling was helpful to distinguish primary lung cancer from metastatic pancreatic adenocarcinoma to the lung. However, loss of DPC4 was the only reliable marker because mucinous bronchioloalveolar carcinomas were frequently negative for thyroid transcription factor 1.39
The distinction of metastatic pancreatic carcinomas in the ovary from primary ovarian mucinous tumors (atypical proliferative/borderline and carcinoma) can also be difficult because of similarities in morphology.40 Pancreatic carcinomas are notorious for simulating primary ovarian mucinous carcinomas by displaying a multicystic appearance with relatively bland mucinous epithelium simulating a mucinous cystadenoma or an atypical proliferative (borderline) mucinous tumor41 (Figure 6). Seidman et al41 proposed the simple rule that classifies all bilateral mucinous carcinomas as metastasis, unilateral mucinous carcinoma less than 10 cm as metastatic, and unilateral mucinous carcinomas more than 10 cm as primary. However, metastatic pancreatic carcinomas are an exception in that they have a tendency for large growth and cystic change. Thus, features that suggest metastatic pancreatic cancer specifically rather than an ovarian primary include bilaterality, surface desmoplastic implants, size more than 10 cm, and extraovarian spread. Similar to other target organ sites, routine immunolabeling can also aid in diagnosis. Ji et al42 reported that loss of DPC4 labeling is useful when used in an immunolabeling panel for distinguishing metastatic pancreatic carcinomas in the ovary from both primary ovarian mucinous tumors and metastatic mucinous carcinomas derived from other sites. More recently, Cao et al43 found that an immunohistochemical panel including DPC4, mesothelin, fascin, and prostate stem cell antigen is useful for evaluating difficult mucinous tumors in the ovary when the differential diagnosis includes metastatic pancreatic adenocarcinoma.
The last decade has seen a dramatic increase in our understanding of the molecular biology of pancreatic cancer.5 Genome-based studies of pancreatic cancer have identified high rates of allelic loss of chromosomes 9p, 18q, 17p, 1p, and 6q in a series of primary pancreatic cancer xenografts.44,45 The targets of these losses include the tumor suppressor genes p16/CDKN2A on chromosome 9p,46 ,DPC4 (SMAD4, MADH4) on chromosome 18q,47 ,TP53 and MKK4 on chromosome 17p,48,49 and BRCA2 on chromosome 13q.50 The identification of noninvasive precursor lesions of invasive pancreatic ductal adenocarcinoma, pancreatic intraepithelial neoplasia, represents a significant advance in our understanding of the earliest forms of this disease.51 The genetic alterations identified in primary pancreatic cancer and its precursors are now thought to follow a genetic progression model associated with the accumulation of multiple genetic alterations, including telomere shortening52; activating point mutations in the KRAS2 gene; and inactivation of the p16/CDKN2A, TP53, SMAD4/DPC4, and BRCA2 tumor suppressor genes51 (Figure 7). Additional genes targeted in pancreatic cancer have been identified using high-resolution global analysis technologies.53–57 For example, novel candidate oncogenes include SMURF1 on 7q21, FGFR1 on 8p12, BIRC2 and BIRC3 on 11q22, and PAK4 on 19q13, whereas novel candidate tumor suppressor genes include TUSC3 on 8p22 and FEZ1 on 8p23.54–58
Most major molecular studies of pancreatic cancer have only used tissues collected from resection specimens8,51,59–61; thus, the molecular features of advanced stage disease are relatively unknown. Nonetheless, in those studies that have focused on advanced stage pancreatic cancer, differences have been found indicating there is additional information to be gleaned.
Activation of the KRAS2 oncogene follows point mutation of codon 12 and, less often, of codons 13 and 61. These mutations maintain the protein in its active form independent of external signals resulting in activation of downstream effector pathways such as Raf-1, Rac, Rho, or phospho-inositol-3-kinase.62 ,KRAS2 gene mutations are among the most common mutations in pancreatic cancer and can be found in up to 95% of pancreatic carcinoma tissues.63,64 ,KRAS2 mutations can be found in early pancreatic intraepithelial neoplasia (pancreatic intraepithelial neoplasia 1), and the rate increases with advancing cellular atypia in precursor lesions.65,66 ,KRAS2 mutations seem to be prerequisite for pancreatic carcinogenesis but not metastasis, as these mutations are present universally in both primary and metastatic cancers in the same patients.11
TP53 is the most commonly mutated tumor suppressor gene in all human cancers.67 It is a nuclear phosphoprotein with the ability to bind to specific DNA elements and to activate gene transcription. p53 has a central position in cell cycle regulation through its role in inactivating a variety of genes and interrupting cell proliferation at the G1-S checkpoint. Inactivating mutations in TP53 are found in up to 70% of pancreatic cancers, most often due to loss of heterozygosity coupled with a mutation in the remaining allele.48,68,69 In one of the few studies to assess TP53 status in metastatic pancreatic cancer, Ruggeri et al68 immunohistochemically analyzed p53 expression in primary and metastatic pancreatic cancer and found no difference in the rate of abnormal p53 labeling between each group. Their study also included a comparison of p53 labeling in 17 matched pairs of primary cancer and metastasis that indicated a strong correlation of abnormal labeling between the primary and metastatic sites. By contrast, no correlation was found with tumor grade, stage, or metastatic disease. More recently, Embuscado et al evaluated TP53 in a series of metastatic pancreatic cancer tissues obtained at autopsy and also found that TP53 showed a similar rate of inactivation to that reported in early stage pancreatic cancers.11,48 Overall, these 2 studies lend support to the hypothesis that genetic alteration of TP53 occurs relatively early in pancreatic carcinogenesis prior to the development of metastatic spread.
DPC4 (SMAD4, MADH4) is a tumor suppressor gene that encodes for a critical transcription factor involved in the TGF-β signal pathway. Genetic inactivation of the DPC4 gene occurs in 55% of pancreatic adenocarcinomas, most commonly by homozygous deletions or intragenic mutations combined with loss of heterozygosity.47 The TGF-β signaling pathway is known to induce epithelial-mesenchymal transition,25–27 an event that occurs normally during embryonic development and is thought to be a critical event contributing to tumor invasiveness and metastasis.70 Consistent with this notion, there is a significantly greater frequency of DPC4 loss in metastatic pancreatic cancers compared with early stage cancers or pancreatic intraepithelial neoplasias, and loss of DPC4 protein expression has been shown to correlate with a shortened survival in patients with resectable pancreatic cancers.11,71–73 The mechanistic relationship of DPC4 loss and/or TGF-β signaling dysregulation to metastatic propensity of pancreatic cancer is uncertain but may relate to TGF-β–mediated signaling through the PI3K/PTEN pathway.25–27
MKK4 (MP2K4/SEK1) is a member of the mitogen-activated protein kinase family, originally identified as a kinase involved in the stress-activated protein kinase pathway by directly phosphorylating c-Jun NH2-terminal kinase.74 ,MKK4 genetic inactivation by homozygous deletion or mutation has been observed in a minority of pancreatic carcinomas, implicating deregulation of the stress-activated protein kinase pathway in pancreatic carcinogenesis.49 Xin et al75 evaluated MKK4 protein expression patterns by immunohistochemical labeling in a large series of resected, advanced stage, and metastatic pancreatic cancers and noted that loss of MKK4 immunolabeling was significantly more common in distant metastases than in the primary carcinomas in these same patients. Moreover, among matched primary and metastatic cancer tissues, loss of MKK4 immunolabeling was seen in 11% of primary cancers but in 37% of distant metastases indicating that MKK4 loss correlates with metastasis formation. The MKK4 gene has been reported to function as a metastasis-suppressor gene in ovarian and prostatic carcinomas, in which the loss of MKK4 gene expression facilitates metastasis formation without affecting primary tumor growth.76 However, using a somatic gene knockout of MKK4 in human cancer cells, Cunningham et al77,78 found that the MKK4 null phenotype was markedly deficient in producing experimental metastases compared with wild-type pancreatic cancer cells and rationalized that a gene dosage-dependent phenotype of MKK4 was needed for optimal metastatic spread of pancreatic cancer.
EphA2 is a transmembrane receptor tyrosine kinase that functions in the regulation of cell growth, survival, angiogenesis, and migration.79 EphA2 is overexpressed in a variety of human cancers, and high levels of EphA2 are associated with aggressive disease and poor clinical outcomes.80 In pancreatic cancers, EphA2 protein is also relatively overexpressed in pancreatic adenocarcinoma cell lines with higher metastatic potential and suppression of EphA2 expression appears to attenuate the invasive phenotype of these same cell lines.81,82
Mudali et al83 evaluated the immunohistochemical labeling patterns of EphA2 protein in a large series of matched primary and metastatic pancreatic cancers and also found frequent overexpression of EphA2 in pancreatic carcinomas compared with benign ducts or pancreatic intraepithelial neoplasia lesions. However, they also noted a significant correlation of EphA2 immunolabeling intensity and organ site metastasis, in that relatively low levels of EphA2 labeling were seen in liver metastases, whereas high levels were seen in association with lymph node metastases (Figure 8). The authors hypothesized that in addition to promoting metastasis, EphA2 levels may contribute to organ site–specific metastasis as well. In rare cases, overexpression of EphA2 was due to amplification at chromosome 1p36, providing a potentially genetic basis for enhanced metastatic ability. These findings are exciting as antibody-mediated targeting of EphA2 has been shown to cause EphA2 internalization, degradation, and loss of oncogenic signals, thus highlighting EphA2 as a therapeutic target for advanced stage pancreatic cancer.84
Gene Expression Profiling
Global expression platforms, such as complementary DNA or oligonucleotide microarrays and serial analysis of gene expression, have helped elucidate large numbers of differentially expressed transcripts in pancreatic cancer. However, most studies have largely focused on the differential gene expression between normal pancreatic tissue and primary pancreatic adenocarcinoma and thus have not addressed pancreatic cancer progression and metastasis specifically.61,85–89
Campagna et al90 used gene expression microarrays to evaluate human pancreatic cancer tissues for global changes in expression that accompany metastatic spread and found no reproducible differences between primary cancers and metastases, whether using matched or unmatched samples. However, a comparison of the expression profiles of early stage and late stage primary carcinomas specifically revealed significant differences in gene expression that were most prominent in gene classes that related to the MAPK and Wnt pathways, metabolism, immune regulation, and cell-cell and cell-matrix interactions. Of interest, one of the genes identified and validated was MAX interacting protein 1 (MXI1), a member of the max dimerization protein family of transcriptional repressors that negatively regulates c-myc.91 Amplification and overexpression of c-myc has been well described in pancreatic cancer where it acts as a central regulator of gene transcription and predisposes cells to apoptosis under nutrient, growth factor, or oxygen deprivation conditions.92,93 Thus, although c-myc amplification and/or overexpression may promote pancreatic carcinogenesis,94 upregulation of MXI1 may occur during disease progression to refine c-myc oncogenic signals and to protect cancer cells from c-myc–induced apoptosis.93 This expression profile associated with growth at the primary site did not overlap with that found when comparing normal pancreatic tissues to pancreatic cancers, leading Campagna et al90 to suggest that in addition to the robust changes in expression that accompany pancreatic carcinogenesis, specific and focused changes occur in association with growth at the primary site that may promote the outgrowth of metastatic clones.
Due to the lack of available high-quality metastatic cancer tissues for expression profiling, a variety of studies have addressed pancreatic cancer metastasis using in vivo models of spontaneous liver metastasis from orthotopic xenografts.95–98 Although all have demonstrated significant differences in gene expression among the primary and metastatic sites, there is little overlap among different gene sets to suggest a commonly dysregulated pathway in pancreatic cancer metastasis formation. For example, Niedergethmann et al98 performed transcriptional profiling of orthotopic xenografts and identified 25 statistically significant upregulated and 181 downregulated genes when comparing the liver metastases with the primary orthotopic tumor. Eight of these genes (PAI-1, BNIP3l, VEGF, NSE, RGS4, HSP27, GADD45A, PTPN14) were validated in a semiquantitative immunohistochemical analysis, and overexpression analyses revealed 66 significantly dysregulated pathways associated with cell proliferation, cell stress, cell communication, and metabolic and cytokine function. Suemizu et al96 also identified statistically significant differences in gene expression between orthotopic xenografts and their spontaneous metastases. However, in their study, only S100A4 expression correlated with the ability to form liver metastases, as evaluated in a quantitative model of metastasis in NOG (NOD/Shi-scid IL2r γ null) mice.
Nomura et al99 described a robust system in which they established HPC-1H5, a highly liver metastatic cell line, and HPC-1P5a, a highly peritoneal disseminating cell line, from the parental pancreatic cancer cell line HPC-1. Using expression profiling to interrogate the gene expression of each line, they found nonoverlapping sets of dysregulated genes in the hepatic and peritoneal disseminating cell lines. For example, several integrins such as α1 integrin (CD49A), α2 integrin (CD49B), α2b integrin (CD41), and β1 integrin (CD29) were upregulated in the liver metastasizing cell line, whereas angiopoetin 1 and MMP10 were upregulated genes of the peritoneal disseminating cell line.
Hedgehog Signaling Pathway
Hedgehog expression is normally restricted to embryogenesis and is instrumental in patterning tissues and structures of the early embryo, including the pancreas.100,101 After pancreatic development, Hedgehog signaling is restricted to the maintenance of stem cell pools.102,103 In normal Hedgehog pathway function, the transporter-like protein Patched acts catalytically to inhibit activation of the 7 transmembrane protein Smoothened. Patched activity is blocked by binding of Hedgehog ligand, thus liberating Smoothened for activation of transcriptional targets mediated by the Gli family of latent transcriptional factors (Figure 9).103
Aberrant reactivation of the Hedgehog signaling pathways, and concomitant overexpression of its target genes, is a common feature of pancreatic cancer.104,105 Hedgehog signaling has also been implicated in pancreatic cancer metastasis. Feldmann et al106 reported that Hedgehog target gene GLI1 levels are upregulated in tissue samples of metastatic human pancreatic cancer samples compared with matched primary tumors. Moreover, inhibition of Hedgehog signaling with the small molecular cyclopamine caused a profound decrease in metastatic spread in this orthotopic xenograft model. A combination of gemcitabine and cyclopamine completely abrogated metastases in this same model while significantly reducing the size of the orthotopic tumors, suggesting a central role of the Hedgehog pathway in pancreatic cancer metastasis.
Metastatic spread from pancreatic cancer remains a major hurdle in combating this disease. A better understanding of the clinical, genetic, and molecular features of metastatic pancreatic cancer will afford novel areas for therapeutic intervention.
The authors have no relevant financial interest in the products or companies described in this article.
Reprints: Christine A. Iacobuzio-Donahue, MD, PhD, The Johns Hopkins Medical Institutions, Department of Pathology, Division of Gastrointestinal/Liver Pathology, 1550 Orleans St, CRB2, Room 343, Baltimore, MD 21231 (firstname.lastname@example.org)