Pancreatic cancer differentiation plasticity

The specific cell of origin responsible for generating pancreatic intraepithelial neoplasia and pancreatic ductal adenocarcinoma remains unknown. During development, epithelial stem cells within embryonic pancreatic epithelium give rise to mature acinar, ductal, and islet elements. Emerging evidence suggests that cells with precursor potential also exist within adult pancreas, resulting in significant developmental plasticity among both endocrine and exocrine cell types. In this review, the contribution of developmental plasticity in initiating pancreatic metaplasia and neoplasia is considered, and evidence supporting a role for epithelial stem cells in pancreatic cancer is discussed.

The embryonic pancreas is first apparent as dorsal and ventral evaginations of prepatterned foregut endoderm, evident by day E9.5 in the mouse. As suggested by Edlund, the complex events required for normal pancreatic development may be considered as three resolvable components. First, foregut endoderm becomes patterned to form dorsal and ventral pancreatic buds. Second, cells undergo lineage commitment to either endocrine or exocrine cell fates. Third, pancreatic morphogenesis occurs by way of extensive growth and branching. During this process, islet, acinar and ductal cell types differentiate from common precursor cells within the developing pancreatic epithelium. Endocrine differentiation appears to precede commitment to exocrine cell fates, with transcripts for somatostatin detectable as early as E7.5 in the mouse, and transcripts for glucagon and insulin detectable by E8.5–E9.0. In contrast, acinar-specific markers such as carboxypeptidase A and amylase are not apparent until day E10.5–E12.5, and expression of the duct-specific marker carbonic anhydrase is not apparent until E11.5. Expression of these exocrine-specific markers occurs well before the onset of morphologic differentiation, with histologically distinct acinar and ductal ceils not appearing until around E15.

The mechanisms regulating lineage commitment in the exocrine pancreas have recently been reviewed. In order for normal exocrine pancreatic differentiation to occur, a subpopulation of precursor cells must avoid undergoing early endocrine differentiation; maintenance of this “reserve” precursor population requires active Notch signaling, thereby preventing expression of Ngn3 and other basic helix-loop-helix (bHLH) transcription factors associated with endocrine differentiation. Normal exocrine differentiation also requires permissive signals provided by adjacent pancreatic mesenchyme, including laminin-1 and soluble follistatin. The relative balance between endocrine, acinar, and ductal cell types is further determined by soluble growth factors, including activin, TGF-ß1, EGF receptor ligands, and FGF-1, -7, and -10.

As previously reviewed, the effects of these signaling pathways on differentiation of endocrine and exocrine pancreatic cell types appear to be conveyed by a hierarchy of lineage-specifying transcription factors. Many of these factors are required not only for the regulated expression of endocrine-and exocrine-specific gene products in adult pancreas, but also for normal morphogenesis and cellular differentiation during pancreatic development. Among these, the homeodomain transcription factor Pdx1 and the paired/homeodomain factor Pax6 may have special relevance in the initiation of pancreatic ductal neoplasia.

The mammalian Pdx1 gene represents the mammalian homologue of the Xenopus XlHbox8 gene. Also referred to as IPF-1 (insulin promoter factor-1), STF-1 (somatostatin transcription factor-1), and IDX-1 (islet-duodenum homeobox gene-1), this gene was first characterized as an important regulator of insulin and somatostatin gene transcription. In addition to its role in regulating endocrine gene expression in adult pancreas, Pdx1 is also required for normal pancreatic development. Homozygous deletions result in the successful specification of pancreatic mesenchyme and initial pancreatic bud formation, but severely aborted morphogenesis and failure to generate differentiated islet, acinar, or ductal cell types. Elegant recombination experiments have confirmed that the morphogenetic defect in Pdx1—/— pancreas resides in the epithelium, with Pdx1—/— mesenchyme fully able to support growth of wild type epithelium but Pdx1—/— epithelium unable to respond to wild type mesenchymal signals.[ 18]

Consistent with its required role in early pancreatic morphogenesis, analysis of Pdx1 gene activation during murine pancreatic development confirms expression by pluripotent stem cells within the embryonic ductal epithelium which ultimately give rise to mature islet, acinar, and ductal elements. Beginning on day E8.5, Pdx1 is expressed in a segment of prepatterned epithelium within the embryonic foregut; the appearance of Pdx1 precedes the initial detection of insulin and glucagon one day later. Pdx1 expression is subsequently observed in the majority of epithelial cells within the dorsal and ventral pancreatic buds. Expression by pancreatic duct cells remains relatively uniform up to day E15.5, but subsequently declines. Co-expression with amylase is also initially observed in developing acinar cells. By E17.5, however, the majority of pancreatic duct and acinar cells show down-regulated Pdx1 expression, with high levels of Pdx1 increasingly restricted to developing islets. Co-localization with glucagon is initially observed in a subset of early glucagon-expressing cells, but no Pdx1 +/glucagon+ cells persist past day E14.5. In contrast, co-expression with insulin is maintained throughout embryonic development, with most adult ß cells continuing to express Pdx1.

In contrast to the profound disruption in pancreatic morphogenesis observed in the absence of Pdx1, the defects resulting from Pax6 deletion appear to be restricted to developing endocrine cells. Pax genes are a highly conserved family of developmental control genes that encode nuclear transcription factors containing the paired box DNA-binding domain. Within this family, the Pax3, Pax4, Pax6, and Pax7 genes contain a homeobox DNA-binding domain in addition to the paired domain. In vertebrates, Pax genes have been implicated as key regulators during organogenesis of kidney, ear, eye, nose, vertebral column, brain, and pancreas. As a group, these transcription factors appear to play a role in mediating epithelial-mesenchymal interactions and stimulating cellular proliferation. A role in preventing apoptosis has also been suggested. The important role of these genes in regulating cell proliferation is further supported by the fact that overexpression of either Pax1, Pax2, Pax3, Pax6, or Pax8 results in malignant transformation of NIH3T3 cells.

In pancreas, Pax6 is initially expressed on day E9 by a subset of cells in the primitive pancreatic epithelium, and later becomes restricted to mature endocrine cells. The Pax6 protein binds a common element in the glucagon, insulin, and somatostatin promoters, conferring transactivation. Spontaneous or targeted deletions of Pax6 result in disorganized islet formation, as well as significant decreases in islet cell numbers. The development of mature of cells may be particularly affected. When Pax6 deletion is combined with targeted disruption of the Pax4 gene, a complete failure to develop mature islet cells is observed.

In addition to these two transcription factors, which in adult pancreas are expressed primarily by endocrine cells, the exocrine-specific bHLH factor p48 also acts as an important regulator of pancreatic epithelial differentiation. The p48 component of the heterotrimeric PTF1 transcription factor complex is the only factor so far identified to be specifically expressed in exocrine tissue and also to be required for normal exocrine differentiation. PTF1 was originally identified based upon its ability to bind conserved sequence elements in the 5' flanking region of exocrine-specific genes. The heterotrimeric PTF1 complex includes E12, HEB/REB/Alf1 and p48. Among these, p48 represents the only cell-specific component. p48 is a bHLH transcription factor that binds to canonical CANNTG E-box sites. Antisense inhibition of p48 expression in exocrine cell lines demonstrates that ongoing p48 activity is required for maintenance of exocrine differentiation. Evidence that p48 is also required for normal development of the exocrine pancreas has been provided by analysis of p48—/— mice, in which a complete absence of acinar tissue is observed. Endocrine differentiation appears to proceed in a relatively normal manner in these mice, with all four islet cell types initially detectable in close association with duct-like structures. As development progresses, however, islet cells appear to continue abnormal migration through the dorsal mesentery, ultimately coming to rest within the spleen.
DEVELOPMENTAL PLASTICITY IN ADULT PANCREAS

Based upon the common derivation of islet, acinar and ductal cell types from precursors within embryonic pancreatic epithelium, it is perhaps not surprising that the adult pancreas demonstrates considerable developmental plasticity. In fact, recent evidence suggests a dramatic capacity for metaplastic interconversion between different pancreatic cell types. In reviewing this evidence, it is important to consider potential mechanisms for metaplasia, defined as the replacement of one predominant cell type within a tissue by an alternate cell type. As suggested by Bouwens, metaplasia in adult pancreatic tissue may result from conversion of fully differentiated epithelial cells from one cell type to another (transdifferentiation hypothesis) or, alternatively, by changes in the frequency with which undifferentiated precursor cells commit to either islet, acinar or ductal lineages (stem cell hypothesis). While the specific examples of pancreatic metaplasia discussed below may argue either for or against these competing hypotheses, it is important to note that proof of in vivo transdifferentiation remains elusive, and that the presence, location, and identity of undifferentiated precursor cells in adult pancreas have only recently been established.

Specific examples of developmental plasticity in adult pancreas include ductal-to-islet metaplasia (islet neogenesis), duct generation within islets, and acinar-to-ductal metaplasia. In many cases, these events appear to be initiated by cellular injury, and may therefore be considered regenerative responses. For example, following either immunologic or pharmacologic injury to pancreatic ß-cells, an increase in endocrine cell mass is noted within ductal epithelium, with subsequent initiation of islet neogenesis. Similarly, acinar cell injury induced by pancreatic duct ligation appears to represent an initiating event in acinar-to-ductal metaplasia. The relationship between acinar cell injury and metaplastic expansion of ductal epithelium has recently been clarified by examination of histologic responses to duct ligation in wild-type mice as well as mice with homozygous deletions in the p53 tumor suppressor gene. In wild type mice, pancreatic duct ligation results in widespread p53 activation, acinar cell apoptosis, cellular inflammation, and expansion of proliferating ductal epithelium (acinar-to-ductal metaplasia). In contrast, up-regulation of p53 target genes and acinar cell apoptosis doe not occur following duct ligation in p53—/— mice. When acinar cell mass is preserved in this manner, expansion of ductal epithelium is also prevented, confirming that acinar-to-ductal metaplasia represents a regenerative response to acinar cell injury in this model.

While the precise mechanisms for these metaplastic events remain uncertain, these findings confirm that adult pancreatic tissue is characterized by significant developmental plasticity, and suggest that developmental programs responsible for normal embryonic differentiation and morphogenesis may become reactivated in adult pancreatic tissue. This concept is reinforced by analysis of Pdx1 gene expression during in vitro acinar-to-ductal metaplasia. Using a primary culture system involving rat pancreatic acini, Rooman and colleagues have demonstrated spontaneous acinar-to-ductal conversion characterized by loss of amylase expression and acquisition of ductal markers including cytokeratin-7 and -20. During this metaplastic conversion, Pdx1 gene activation was documented by increases in both Pdx1 transcript and Pdx1 protein, similar to that observed in embryonic pancreatic epithelium.
DEVELOPMENTAL PLASTICITY IN HUMAN PANCREATIC NEOPLASIA

Not surprisingly, the developmental plasticity observed in normal pancreatic epithelium is also observed in neoplastic pancreas. Although 90% of all pancreatic neoplasms may be classified as ductal adenocarcinoma, the cell populations responsible for generating pancreatic cancer have not been well characterized. Pancreatic ductal cancers arise infrequently within the main pancreatic duct, and precursor lesions have only recently been defined. Further uncertainty regarding the origin of these lesions is generated by expression of a complex pattern of lineage markers in neoplasms exhibiting predominantly ductal morphology. Kim and colleagues examined a series of primary and metastatic pancreatic ductal adenocarcinomas using immunohistochemistry for a battery of ductal, acinar, and islet lineage markers. Among 34 tumors, 15 (44%) demonstrated a pluripotent differentiation capacity as evidenced by combined expression of either ductal and acinar markers (N = 3), ductal and islet markers (N = 8), or markers of all three lineages (N = 4). In this series, cell lineage ambiguities tended to appear most frequently in poorly differentiated lesions.

Additional studies have suggested that the differentiation capacities of pancreatic ductal adenocarcinoma cells may extend even beyond pancreas-specific cell fates. Among 88 pancreatic ductal adenocarcinomas studied by Sessa and colleagues, the majority expressed antigens normally found in gastric and/or intestinal epithelia. While all 88 of these lesions expressed one or more markers of normal pancreatic ductal epithelium such as DU-PAN-2, all but 2 of the 88 expressed at least one gastric or intestinal antigen, and most expressed more than one. These markers included M1 (93%) and cathepsin E (92%), markers of gastric surface-foveolar cells; pepsinogen II (51%), a marker of gastroduodenal mucopeptic cells; CAR-5 (48%), a marker of colorectal epithelial cells; and M3SI (35%), a marker of small intestinal goblet cells. These expression profiles correlated with ultrastructural features consistent with gastric foveolar cells, gastric mucopeptic cells, intestinal goblet cells, intestinal columnar cells, as well as pancreatic ductal epithelium.

This pluripotent differentiation capacity likely reflects the common derivation of gastric, pancreatic, and intestinal epithelial cells from precursor cells in foregut endoderm. In rare cases, this capacity may also include the potential for hepatoid differentiation, recalling the common derivation of liver and ventral pancreas from ventral foregut endoderm.

Together, these findings confirm the pluripotent differentiation capacity of pancreatic ductal adenocarcinoma, but fail to resolve questions regarding the “cell of origin” for this tumor. Specifically, it remains uncertain whether cell lineage ambiguities in pancreatic cancer reflect the differentiation capacities of an originating epithelial stem cell, or whether these divergent cell fates reflect plasticity of mature, differentiated pancreatic cell types.
PANCREATIC INTRAEPITHELIAL NEOPLASIA: A PRECURSOR FOR PANCREATIC CANCER

Recent molecular and histologic studies have further suggested that foci of ductal proliferation may represent precursor lesions for pancreatic cancer. These lesions are frequently observed in the setting of chronic pancreatitis. Loss of acinar cells, proliferation of small ductal complexes and pancreatic ductal hyperplasia are common features of chronic pancreatitis, a condition associated with a 16-fold increased risk for eventual pancreatic cancer development. Similar areas are frequently observed in pancreatic tissue adjacent to areas of invasive pancreatic cancer.

In an attempt to achieve uniformity in nomenclature regarding these lesions, the term pancreatic intraepithelial neoplasia (PanIN) has been proposed. Further evidence in support of PanIN lesions as precursor lesions for invasive ductal adenocarcinoma is provided by analysis of molecular changes associated with PanIN lesions of increasing severity. These studies have confirmed the progressive accumulation of genetic changes characteristic of invasive pancreatic adenocarcinoma in PanIN's of increasing severity, with activating mutations in K-ras occurring in low-grade lesions, and progressive loss of p16, p53, and DPC4 in later lesions. Together with evidence demonstrating the initial presence of PanIN lesions in patients who later developed infiltrating carcinoma, these findings strongly support a precursor role for PanIN lesions.

The mechanisms of PanIN lesion formation remains poorly understood. Among the possible origins for these lesions, at least three possibilities have been considered. First, PanIN lesions may result from proliferation and progressive dysplasia of fully differentiated duct cells within the branching network of pancreatic ductal epithelium. Such a mechanism would appear likely in the generation of intraductal papillary mutinous neoplasia (IPMN), which frequently affects large pancreatic ducts. Alternatively, PanIN lesions may be generated by metaplastic conversion of other pancreatic cell types through the process of either islet-to-ductal metaplasia or acinar-to-ductal metaplasia. Considerable experimental evidence exists in support of both forms of metaplasia as relevant events in pancreatic tumorigenesis.
Islet-to-Ductal Metaplasia as an Initiating Event in Pancreatic Neoplasia

Several studies have demonstrated the potential for apparent generation of ductal epithelium by islet tissue. When cultured under appropriate conditions, isolated islets harvested from either hamster or human pancreas undergo transition to a duct-like epithelium characterized by loss of islet hormone expression and expression of the ductal markers cytokeratin-19 and CA19-9. As in the case of all studies to date involving metaplastic transition between pancreatic cell types, it is not clear whether these metaplastic duct cells arise from an undifferentiated precursor population associated with pancreatic islets, or by way of islet cell transdifferentiation.

Islet-to-ductal metaplasia has also been observed in vivo following transgenic expression of keratinocyte growth factor (KGF/FGF7) under control of the human insulin promoter. These mice demonstrated markedly abnormal patterns of islet differentiation, with some islets undergoing islet-to-hepatocyte metaplasia, and other islets undergoing islet-to-ductal metaplasia. Intra-islet duct complexes developing in these mice showed high rates of bromodeoxyuridine (BrdU) incorporation, and also expressed the ductal marker carbonic anhydrase II.

Based upon this potential for islet-associated cells to generate metaplastic ductal epithelium, it is not surprising that these cells may also give rise to ductal neoplasia under certain experimental conditions. In one study, pancreatic islets isolated from juvenile mice were infected with a retrovirus carrying polyomavirus middle T-antigen. Cell lines generated from these transformed islets expressed both cytokeratin, a marker of ductal epithelium, as well as a limited array of islet markers, including somatostatin and pancreatic polypeptide. When middle T-antigen transformed cells were inoculated into mice, they rapidly formed well-differentiated ductal adenocarcinomas that expressed cytokeratin but not islet cell markers. Transformed cell lines capable of forming similar cytokeratin-positive tumors with glandular elements have also been generated following treatment of isolated hamster islets with the carcinogen, N-nitrosobis( 2-oxopropyl)amine (BOP). Together, these studies suggest that precursor cells capable of generating pancreatic ductal adenocarcinoma reside within preparations of isolated pancreatic islets.

When BOP is administered in vivo to Syrian golden hamsters, pancreatic ductal adenocarcinomas are generated that demonstrate strong similarity to human tumors, including ductal morphology, tendency for peritoneal and hepatic metastasis, and high rates of activating K-ras mutations. Pour and colleagues have generated considerable data suggesting that the cellular targets for carcinogen transformation in this model reside within or adjacent to pancreatic islets. The initial histologic manifestation of pancreatic tumorigenesis in this model appears to involve the appearance of intra-islet ductal structures (PanINs), which undergo progressive hyperplasia and eventual malignant transformation. Destruction of islet cells by administration of streptozotocin or alloxan inhibits tumor formation in this model, as does prevention of islet cell proliferation by the anti-gluconeogenic compound metformin. In contrast, manipulations which stimulate islet cell proliferation enhance pancreatic ductal carcinogenesis in this model. Further support for the islet derivation of BOP-induced ductal tumors in the hamster is provided by studies involving transplant of isolated islets from male hamsters into the submandibular gland of female hamsters, followed by in vitro administration of BOP. In this study, female hamsters receiving submandibular transplants of male islets (but not immortalized pancreatic duct cells) developed PanIN-like lesions and eventual ductal adenocarcinomas expressing a Y-chromosome marker within the recipient salivary gland.
Acinar-to-Ductal Metaplasia as an Initiating Event in Pancreatic Neoplasia

As an alternate to an islet origin of PanIN lesions, the frequently associated loss of acinar cells observed adjacent to these precursors suggests that PanINs may arise through a process of acinar-to-ductal metaplasia. When the EGF receptor ligand TGF-a is targeted to mouse pancreas using acinar cell-specific elastase-1 promoter/ enhancer elements (E1-TGF-a transgenics), extensive acinar-to-ductal metaplasia is observed. Further evaluation of E1-TGF-a transgenic mice has demonstrated accumulation of increasingly dysplastic PanIN lesions, with eventual generation of malignant ductal adenocarcinoma. When bred onto a p53 +/- background, these animals develop invasive, metastatic ductal adenocarcinoma characterized by ras activation, loss of heterozygosity for the remaining p53 allele, frequent biallelic inactivation of p16, and occasional loss of heterozygosity for DPC.[ These results confirm the premalignant nature of acinar-to-ductal metaplasia and suggest that pancreatic adenocarcinomas arising from metaplastic ductal epithelium share many features in common with human pancreatic cancer.
Similarities Between Premalignant and Embryonic Pancreatic Epithelium

In order to gain further insight into the nature of premalignant metaplastic ductal epithelium, we have further examined patterns of gene expression during TGF-a-induced acinar-to-ductal metaplasia. Using a battery of lineage markers including CFTR, amylase, chromogranin A, insulin, and glucagon, the pluripotent differentiation capacity of this epithelium is revealed. Specifically, premalignant metaplastic epithelium demonstrates widespread evidence of ductal differentiation as indicated by CFTR expression, but also contains areas characterized by acinar differentiation as indicated by typical acinar morphology and ongoing expression or amylase. In addition, premalignant metaplastic epithelium demonstrated foci of active islet neogenesis, characterized by expression of chromogranin A, glucagon, and insulin.  In this regard, the pluripotent differentiation capacity of metaplastic ductal epithelium shows remarkable similarity to the epithelium of the embryonic pancreas.

Further support for a significant analogy between metaplastic epithelium and embryonic pancreatic epithelium is provided by examination of the homeodomain transcription factors Pdx1 and Pax6. Expression analysis of these genes has been facilitated by insertion of a lacZ reporter into one allele of either gene, generating Pdx1lacZ/+ or Pax6lacZ/+ mice. Using in vivo reporter gene analysis in bi-transgenic Pdx1lacZ/+/MT-TGF-a mice, it is apparent that TGF-a-induced metaplastic epithelium is characterized by widespread Pdx1 gene activation. Similar to the pattern of differentiation observed in embryonic pancreas, the majority of Pdx1-positive metaplastic duct cells do not express insulin, except in focal areas of islet cell neogenesis.

Using a similar approach, we have examined the pattern of Pax6 expression in Pax6lacZ/+/ wild-type and Pax6lacZ/+/MT-TGF-a bi-transgenic mice. As previously reported,[ 26] Pax6 gene activation was restricted to islet cells in wild-type mice lacking the TGF-a transgene, with no Pax6 expression observed in normal ductal epithelium. In contrast, metaplastic epithelium in Pax6lacZ/+/MT-TGF-a mice demonstrated focal activation of the Pax6lacZ/+ allele, consistent with previous observations regarding expression of Pax6 protein in this epithelium. Together, these findings suggest that PanIN-like metaplastic epithelium displays features characteristic of embryonic pancreas, including a high proliferative rate, pluripotent differentiation capacity, and an embryonic pattern of homeodomain transcription factor expression.

Given the fact that Pax6 may act as a transforming oncogene when expressed outside of its normal developmental context, these findings suggest that metaplastic reversion to an embryonic-like epithelium may play a role in the initiation of pancreatic ductal adenocarcinoma. Further support of this theory has recently been provided by analysis of transgenic mice expressing Pax6 under regulation of Pdx1 promoter elements. These mice exhibit abnormal ß-cell maturation, increased ß-cell apoptosis, and diabetes. In addition, pancreatic tissue from these mice demonstrated hyperplasia of both islet cells and ductal epithelium, with eventual formation of tumors resembling ductal cystadenomas.

The adult pancreas exhibits remarkable developmental plasticity, reflecting the origin of acinar, ductal and islet cell types from common precursor populations in embryonic pancreatic epithelium. During normal pancreatic development, lineage commitment and cellular differentiation are regulated by cell-cell contact, epithelial-mesenchymal signaling, soluble growth factors, and lineage-specific transcription factors. Similar pathways appear to be involved in the regulation of cellular differentiation in adult pancreas, including the altered differentiation pathways activated during various forms of pancreatic metaplasia. Just as normal adult pancreas exhibits considerable capacity for metaplastic interconversion between predominant cell types, so does neoplastic pancreatic epithelium frequently demonstrate a multipotent differentiation capacity. This capacity may be at least partly conferred through participation of undifferentiated precursor cells in early pancreatic neoplasia. Recent evidence suggests that PanIN-like duct lesions may be generated from either endocrine or exocrine pancreatic tissue, and that these metaplastic conversions may recapitulate several components of normal pancreatic development. The recent identification of precursor cells distributed within both exocrine and endocrine tissue[ suggests a candidate cell type which may be responsible for these events.

Last updated Jan 2/07