Molecular Genetics of Pancreatic Ductal Adenocarcinomas and Recent Implications for Translational Efforts

Genomic (DNA) Alterations

Genomic DNA abnormalities frequently found in pancreatic cancers comprise chromosomal aberrations, copy number changes, activating mutations of oncogenes, as well as specific silencing mutations of tumor suppressor and caretaker genes, epigenetic silencing, telomeric alterations, and mutations in mitochondrial DNA (mtDNA).

Copy Number Aberrations

Due to chromosomal instability, a common feature of most solid tumors, almost every pancreatic cancer harbors several numerical or structural chromosomal alterations revealed by cytogenetic analysis.^4,5 The most common numerical changes observed in pancreatic cancer are losses on chromosomes 6, 12, 13, and 18, as well as gains on chromosomes 7 and 20; chromosomal breaks and rearrangements most frequently occur in regions involving 1p, 1q, 3p, 6q, 7q, 11p, 17p, and 19q.

Another technique that is commonly applied to identify regions of genomic losses at a “higher resolution” by means of polymorphic microsatellite markers is called allelotyping. In a recent study Iacobuzio-Donahue et al analyzed ∼80 pancreatic cancer xenografts by means of 386 microsatellite markers.⁶ Allelic losses found most commonly involved chromosomal regions 9p, 17p, and 18q, covering the tumor suppressor genes CDKN2A, TP53, and DPC4/SMAD4/MADH4, respectively. Additional losses were frequently found in 3p, 4q, 5q, 6q, 8p, 12q, 14q, 21q, and 22q. Many of these regions have been linked to candidate tumor suppressor genes, eg, the stress-activated protein kinase MKK4 (17p), which has been suggested to be involved in metastatic spread,⁷ or receptors of the transforming growth factor (TGF)-β signaling pathway TGFBR1 (9q), TGFBR2 (3p),⁸ and ACVR1B (12q).⁹ Interestingly, allelotype analysis of microdissected pancreatic intraepithelial neoplasia (PanIN) samples revealed loss of heterozygosity in several of the chromosomal regions also found in pancreatic cancer, including 9p, 17p, and 18q.^10,11

Comparative genomic hybridization (CGH) can be used to discover genomic deletions as well as amplifications, providing the potential to uncover both potential tumor suppressor genes and oncogenes. For CGH, samples of non-neoplastic and tumor cell DNA are labeled with different dyes and hybridized against each other. Subsequently, the relative ratio of the two dyes indicates regions of cancer-associated gains or losses. Conventional CGH was originally performed using metaphase spreads, with the major drawbacks of relatively low resolution and frequent difficulties to map precisely the regions of genomic amplifications or losses.¹² More recently, several array-based CGH techniques have been developed, using microarrays that are spotted with bacterial artificial chromosomes (BAC arrays),¹³ cDNAs (cDNA microarrays),¹⁴ or stretches of oligonucleotides (representational oligonucleotide microarray analysis),¹⁵ providing a significantly higher resolution than conventional CGH (up to 30 kb) and often allowing for precise mapping of deleted or amplified regions and genes included therein. In the setting of pancreatic cancer, array CGH revealed several genomic amplifications, including C-MYC (8q), EGFR (7p), KRAS22 (12p), AKT2 (19q), and AIB1 (20q), as well as deletions, including DPC4/SMAD4/MADH4 (18q), CDKN2A (9p), FHIT (3p), and MKK4 (17p).^16,17 Using the example of thymidylate synthase, which has previously been linked to responsiveness to 5-fluorouracil treatment, a recent report by Brody et al suggests that results from studies examining copy number aberrations should be interpreted with particular caution, since copy numbers determined in tumor cells are not necessarily identical throughout the whole tumor and might vary over time, in response to chemotherapeutic agents or in metastatic foci as compared to primary tumors.¹⁸

Nuclear DNA Mutations: Oncogenes

Activating point mutations within the KRAS oncogene (12p) are present in 80 to 90% of pancreatic cancers, most commonly affecting codon 12 but also 13 or 61.¹⁹ The activating mutations abolish the intrinsic GTPase activity of KRAS, resulting in constitutive activation of intracellular signal transduction (Figure 1). Of note, activating KRAS mutations are not only the most frequently found genetic abnormalities in pancreatic cancer but also seem to be among the earliest changes observed in nonmalignant precursor-lesions, already being present in about 30% of PanIN-1 lesions.^21,22 Rare pancreatic cancers with wild-type KRAS usually harbor mutations of BRAF.²³ Since both KRAS and BRAF function in activating the same Ras/Raf/MAP kinase signaling pathway, it explains why mutations of these two genes occur in a virtually exclusive pattern and further underscores the extraordinary importance of this signaling pathway in the genesis of pancreatic cancer. Other oncogenes involved in pancreatic cancer include CMYC, AKT2, and EGFR. CMYC amplifications and concomitant overexpression of CMYC can be detected in 50 to 60% of pancreatic cancers,^17,24 providing a potential target for the development of future therapeutic options due to the recent development of compounds specifically inhibiting Myc signaling.²⁵

Although amplifications of AKT2 are found in less than 5% of cases,²⁶ the Akt signaling pathway is activated in 30 to 40% of pancreatic cancers,²⁷ suggesting additional underlying mechanisms of pathway activation, eg, loss of PTEN. Specific inhibitors of Akt signaling have been described,²⁸ which might give rise to novel therapeutic strategies in the near future. Constitutive activation of the EGFR pathway through amplification of the EGFR gene or other mechanisms has been described in pancreatic cancer.¹⁶ Recent approaches were aimed to target this signaling pathway using monoclonal antibodies or small-molecule tyrosine kinase inhibitors.²⁹

Tumor Suppressor Genes

The cyclin-dependent kinase CDKN2A/p16 on chromosome 9p inhibits cell cycle progression through the G₁-S checkpoint and constitutes the most frequently inactivated gene in pancreatic cancers.^19,30 More than 90% of pancreatic cancers show loss of CDKN2A/p16 function, which can occur due to homozygous deletions (∼40%), mutations with loss of the second allele (∼40%), or epigenetic silencing by promoter hypermethylation (10–15%). Interestingly, loss of nuclear p16 protein expression is already observed in 30% of PanIN-1, in 55% of PanIN-2, and in 71% of PanIN-3 lesions.³¹ Approximately 30% of homozygous CDKN2A/p16 mutations also include the MTAP gene, which has recently been proposed as potential therapeutic target.³²

Deleted in pancreatic carcinoma 4 (DPC4/SMAD4/MADH4) on chromosome 18q21 is inactivated in about 55% of pancreatic cancers.³³ This is due to mutations in one allele and loss of the second allele in about 25% of cases and due to homozygous mutations in 30% of cases. Of note, DPC4/SMAD4/MADH4 mutations are very rarely observed in other malignancies.³⁴ Loss of DPC4/SMAD4/MADH4 function results in reduced growth inhibition and increased proliferation through interference with intracellular signaling cascades downstream of cell surface receptors of the TGF-β family (Figure 2). Abrogation of DPC4 function appears to be a rather late event in the pathogenesis of pancreatic carcinomas, as it is not observed in the majority of PanIN-3 lesions.³⁵

Third, TP53 on chromosome 17p is inactivated in 50 to 75% of pancreatic cancers, almost exclusively due to intragenic mutations of one allele in combination with loss of the second allele. Wild-type TP53 induces cell cycle arrest and apoptosis in response to DNA damage, and therefore loss of TP53 function allows for accumulation of additional genetic aberrations. Nuclear accumulation of mutated TP53 is only observed in advanced PanIN-3 lesions, and loss of TP53 function is, like DPC4, considered to be a late event in the pathogenic cascade of pancreatic cancer development.³⁵

While loss of these three (CDKN2A, DPC4 and TP53) genes is found in the majority of pancreatic cancers, other tumor suppressor genes are inactivated in smaller subsets (<10%) of PDAC, eg, LKB1/STK11 on chromosome 19p,³⁶ TGF-βR1 on 9q, TGF-βR2 on 3p, RB1 on 13q,³⁰ and MKK4 on 17p.³⁷ Interestingly, MKK4 seems to be inactivated specifically in metastatic lesions, suggesting that wild-type MKK4 might have the ability to inhibit development of metastases through a yet unknown mechanism.

Caretaker Genes

As opposed to alterations of oncogenes and tumor suppressor genes, which both drive the neoplastic process by increasing tumor cell numbers through increased tumor cell growth or inhibition of cell death and cell cycle arrest, a third class of genes is commonly involved in the development of malignant neoplasias, such that when mutated, they act in a fundamentally different way. Such genes are commonly referred to as caretakers or stability genes.³⁸ Caretaker genes minimize genetic alterations during DNA replication so that loss of their function can lead to accumulation of additional mutations in different other genes.³⁹

The DNA damage repair genes hMLH1 and hMSH2 are inactivated in a small subset of familial pancreatic cancers, but rarely in sporadic cases,⁴⁰ mostly medullary carcinomas of the pancreas.⁴¹ These tumors constitute a separate entity that is important to recognize, since on the one hand they tend to carry a more favorable prognosis than ductal adenocarcinomas and, on the other hand, they may indicate hereditary nonpolyposis colorectal cancer syndrome, in which case patients might benefit from genetic counseling. Medullary carcinomas of the pancreas show a typical medullary histology, which is characterized by poor differentiation, pushing borders, and syncytial growth pattern.^41,42

Moreover, a subgroup of pancreatic cancers carries mutations in genes of the Fanconi's anemia DNA repair pathway. Mutations affecting the BRCA2 gene have been found in approximately 17% of familial pancreatic cancers,⁴³ and FANCC and FANCG mutations have been described as rare findings in sporadic pancreatic cancers,⁴⁴ all of which are thought to be part of the Fanconi's anemia DNA repair pathway. (See Kennedy and D'Andrea ⁴⁵ and Taniguchi and D'Andrea⁴⁶ for a recent review and for a schematic overview of the Fanconi's anemia pathway.)

Xenograft Models

Research aimed at better pathophysiological understanding and development of novel therapeutic strategies against pancreatic cancer has long been hampered by the lack of suitable small animal models that satisfactorily resemble clinical features of human disease. Traditionally, initial screening of potential new drugs for in vivo efficacy is most commonly performed on subcutaneous (athymic nude or SCID) mouse xenograft models, derived from either human pancreatic cancer cell lines or primary cancer tissue specimens, which accurately mirror genotypic features of the parental tumors and susceptibility or resistance to anticancer agents even after serial transplantation.⁹⁴ While these in vivo models are relatively easy to generate and allow for easy screening of putative new drugs on several xenograft tumor lines, and thus are undoubtedly of immense value for initial pharmacological studies, they also suffer from some fundamental drawbacks.^95,96 First, as these xenogenic transplantation models lack T cell or both B and T cell responses, effects of the tumor environment including immunological effects are poorly mirrored. Second, as these models are heterotopic, they do not reflect the physiological microenvironment in the pancreas, including anatomical conditions and effects of direct tumor invasion into neighboring structures, cytokine and chemokine secretion patterns, vascularization and tumor-induced neoangiogenesis. Third, and related to the first two points, subcutaneous xenografts of pancreatic cancers hardly ever metastasize, so that mechanisms involved in metastatic spread cannot be studied using these models, although this is one of the most prominent features of pancreatic cancer in humans that de facto determines the overall survival.

Therefore, recent years have seen increasing use of orthotopic xenograft models, which are more tedious to generate but are able to overcome many of these shortcomings. They are particularly useful to study drug effects on tumor microenvironment including neoangiogenesis and metastatic spread.^87,97 Tumor cells are either injected directly into the mouse pancreas, often in the form of concentrated tumor cell suspensions containing Matrigel, or alternatively are surgically implanted into the murine pancreas as chunks of primary tumor tissues or xenografts. While the first variant is faster and allows generation of xenografts from cell lines without the need to first grow subcutaneous xenografts, the second technique is more reliable in preventing intraperitoneal leakage of tumor cells after injection and also allows for the use of primary tissue samples, thus bypassing the need for in vitro established cell lines.

Transgenic Mouse Models

As opposed to subcutaneous or orthotopic xenografts, transgenic mouse models possess the potential to mimic human disease in a syngeneic system. While they can obviously not represent the whole spectrum of genetic aberrations observed in human pancreatic cancers, these models are of tremendous value for translational studies as they also include a syngeneic tumor environment as well as a fully intact immune system. The earliest models, described almost two decades ago, used acinar-specific elastase promoter to target oncogene expression to the murine pancreas, resulting in neoplasms of predominantly acinar histogenesis.^98,99

The first mouse model most closely resembling histopathological features of human ductal adenocarcinoma was described in 2003.¹⁰⁰ In this model, expression of oncogenic Kras^G12D is suppressed by combination with Lox-STOP-Lox (LSL) constructs. Repression is released on Cre-mediated excision of the LSL cassettes and subsequent recombination. Targeting of transgene expression to the murine pancreas is achieved by expression of Cre recombinase under the control of pancreas-specific promoters Pdx1 or P48. It is commonly assumed that Pdx1-/P48-double positive cells give rise to virtually all mature cells in the pancreas.¹⁰⁰,¹⁰¹,¹⁰² During embryonic development, expression of Pdx1/PF1 starts around E8.5, and P48/PTF1 expression begins slightly later. All of these mice develop ductal lesions resembling human PanIN lesions, which eventually progress into a fully invasive and metastatic adenocarcinoma phenotype in a small percentage (<10%) of animals. The long latency of 6 to 8 months and low frequency suggest the need of additional genetic alterations, likely including the INK4a-Rb or Arf-p53 pathways, whereas the majority of cells expressing oncogenic Kras^G12D alone might undergo ras-induced senescence and thus fail to accumulate additional hits required to develop a fully malignant cancer phenotype.¹⁰³ Similar to these observations, Grippo et al found multifocal acinar cell hyperplasia in Ela-Kras^G12D mice 1 to 2 months of age. Interestingly, by the age of 6 to 18 months, some of these lesions underwent a process the authors referred to as acinar-to-ductal metaplasia and presented with a more duct-like phenotype, including expression of CK-19.¹⁰⁴

In fact, newer transgenic models have recently been described, in that Kras^G12D expression is directed to the pancreas by means of Cre recombinase under the control of a Pdx1 promoter in combination with inhibition of the INK4A/Arf or p53 pathways, respectively: LSL-Kras^G12D; INK4a/Arf^lox/lox;Pdx1-Cre,¹⁰⁵ LSL-Kras;p53^lox/lox;Pdx1-Cre,¹⁰⁶ and LSL-Kras;Trp53^R172H;Pdx1-Cre.¹⁰⁷ Mice with knockout of the INK4a/Arf locus, resulting in loss of both murine p16 and p19 function, develop poorly differentiated carcinomas very rapidly and start to die before 7 weeks of age,¹⁰⁵ whereas inhibition of p53 function, either by genetic knockout¹⁰⁶ or by introduction of a dominant negative p53 allele,¹⁰⁷ preferentially leads to moderately to well-differentiated adenocarcinomas. Of note, abrogation of either INK4a/Arf or p53 signaling alone in the absence of oncogenic Kras does not lead to the development of pancreatic carcinomas or associated precursor lesions, underscoring the crucial importance of Kras signaling in initiating the cascade of events, eventually culminating in a fully malignant phenotype during pancreatic carcinogenesis.¹⁰⁵

The described models, and especially the latter,¹⁰⁷ to date also represent those recapitulating most closely the clinicopathological characteristics of human pancreatic adenocarcinomas and thus carry enormous potential for future translational studies in providing an excellent platform for preclinical evaluation of novel drugs and other therapeutic approaches. Several signaling pathways shown to be aberrantly activated in human pancreatic cancers are also found to be turned on in the discussed mouse models, including the Hedgehog and Notch signaling pathways. As these pathways can be targeted by using blocking antibodies or small-molecule inhibitors, for example, it is tempting to speculate whether therapeutic regimens exploiting blockade of these pathways might have an effect on survival in these preclinical models and might moreover eventually be translated into applications in a clinical setting.

Using a mouse model of TGF-α-induced pancreatic cancer previously described by Wagner and colleagues,¹⁰⁸ a recent study from the same group found that pancreatic carcinomas occurring in C57BL/6-EL-TGF-alpha;Trp53^−/− mice induced a distinct immune response including secretion of proinflammatory cytokines and occurrence of tumor-specific regulatory T lymphocytes in the host. Surprisingly, tumor-derived cell lines did not form xenograft tumors in immunocompetent mice of the same genetic background, a finding that is yet to be fully understood. From these results the authors conclude that spontaneous tumors arising in this mouse model are recognized by the host immune system and that therefore transgenic models might be more suitable than xenografts to evaluate certain immunotherapeutic regimens in a preclinical setting.¹⁰⁹

An immunotherapeutic approach using tumor vaccination with tumor/dendritic cell fusion supplemented by injection of superantigen staphylococcal enterotoxin B leads to generation of tumor-specific cytotoxic T lymphocytes and increased survival in a previously described transgenic mouse model of acinar cell-type pancreatic cancer overexpressing MUC1,^110,111 which is based on a model originally described by Tevethia et al, in which pancreatic acinar carcinomas are induced by pancreas-specific expression of a transgene containing the N-terminal amino acids 1–127 of large T-cell antigen under the control of THE elastase-1 promoter.¹¹²

In the last few years, several other transgenic mouse models targeting different pathways thought to be involved in pancreatic carcinogenesis, which allow deeper insight in underlying etiological and pathogenetic mechanisms, have been described. Three recent mouse models addressed the role of TGF-β signaling in pancreatic cancer (see Figure 2 for an overview of the TGF-β pathway) by pancreas-specific deletion of either SMAD4^113,114 or TGF-β receptor type II (TGFBR2).¹¹⁵ Interestingly, as observed for interruption of INK4a/Arf and p53 signaling pathways, neither SMAD4 nor TGFBR2 deletion alone is sufficient to induce pancreatic neoplasia. However, when combined with oncogenic Kras^G12D, development of fully malignant carcinomas is enhanced with shorter latencies than observed for Kras^G12D alone. While mice lacking pancreatic TGFBR2 expression in the presence of Kras^G12D all develop well-differentiated adenocarcinomas and die of their disease within 200 days, with a median survival of 59 days,¹¹⁵ loss of SMAD4 in Kras^G12D-expressing pancreata leads to development of premalignant precursor lesions of intraductal papillary mucinous neoplasm or mucinous cystic neoplasm type, which can progress into fully malignant carcinomas. For the latter, median survival of 8 months¹¹⁴ and 7 to 12 weeks¹¹³ have been described. TGF-β signaling has previously been proposed to mediate epithelial-to-mesenchymal transition,¹¹⁶ and in line with this concept, INK4A/Arf-null;Kras^G12D mice with wild-type SMAD4 usually present with poorly differentiated carcinomas, whereas mice that also lack SMAD4 expression, develop mostly well- to moderately differentiated pancreatic adenocarcinomas that express epithelial markers such as E-cadherin or cytokeratin-19.

Studies examining the role of Hedgehog signaling demonstrated that overexpression of the Hedgehog ligand Sonic Hedgehog (SHH) in the pancreas is sufficient to induce formation of PanIN lesions.⁷⁹ Introduction of a dominant-active form of the activating Hedgehog transcription factor GLI2 (CLEG2) leads to formation of poorly differentiated pancreatic carcinomas that lack CK-19 expression in about 30% of PDX1-Cre;CLEG2 mice, which seem to develop without evidence of PanINs as precursor lesions.¹¹⁷ Of note, combined expression of CLEG2 and Kras^G12D in the murine pancreas results in formation of PanIN lesions and pancreatic carcinomas in all studied mice, with shorter latency and dramatically decreased overall survival of only 3 to 8 weeks.

Overexpression of COX-2 in the pancreas under the control of a keratin 5 promoter leads to formation of IPMN- and PanIN-like lesions, enhanced ras signaling, inflammation, and fibrosis in a subset of mice.¹¹⁸ Clerc and colleagues described development of pancreatic carcinomas through acinar-to-ductal transition in three of 20 mice overexpressing CCK2/gastrin receptor in pancreatic acinar cells.¹¹⁹