Acinar cell plasticity and development of pancreatic ductal adenocarcinoma

Abstract

Acinar cells in the adult pancreas show high plasticity and can undergo transdifferentiation to a progenitor-like cell type with ductal characteristics. This process, termed acinar-to-ductal metaplasia (ADM), is an important feature facilitating pancreas regeneration after injury. Data from animal models show that cells that undergo ADM in response to oncogenic signalling are precursors for pancreatic intraepithelial neoplasia lesions, which can further progress to pancreatic ductal adenocarcinoma (PDAC). As human pancreatic adenocarcinoma is often diagnosed at a stage of metastatic disease, understanding the processes that lead to its initiation is important for the discovery of markers for early detection, as well as options that enable an early intervention. Here, the critical determinants of acinar cell plasticity are discussed, in addition to the intracellular and extracellular signalling events that drive acinar cell metaplasia and their contribution to development of PDAC.

Of the adult cell lineages of the pancreas, acinar cells show the highest plasticity^1,2. Pancreatic acinar cells can dedifferentiate or transdifferentiate to an embryonic progenitor phenotype that expresses ductal markers, in a process termed acinar-to-ductal metaplasia (ADM)³. Multiple factors have been implicated in mediating ADM, including KRAS hyperactivity and increased inflammatory signalling^4–7 (FIG. 1). The implication of ADM in the development of pancreatic adenocarcinoma was first demonstrated in mice by transgenic overexpression of transforming growth factor (TGF)-α⁸. ADM was also demonstrated in vitro in 3D cell culture, in which mouse acinar cell clusters, in the presence of internal or external stress signalling, oncogenic KRAS, inflammatory cytokines or growth factors that activate epidermal growth factor receptor (EGFR), spontaneously transdifferentiate into duct-like structures^4,6,9–11. Similar 3D cell culture experiments showed that ADM in human acinar cells can be induced by TGFβ¹².

Figure 1. Acinar cell plasticity and metaplasia to duct-like cells in the adult pancreas.

In response to pancreatic injury, the loss of cell–cell and cell–matrix contacts (contact-mediated signalling), loss of polarity, KRAS hyperactivity and increased inflammatory signalling can drive acinar cells to undergo dedifferentiation and transdifferentiation to a duct-like phenotype that is needed for pancreatic regeneration. Acinar-to-ductal metaplasia becomes irreversible in the presence of an oncogenic Kras mutation and persistent growth factor signalling, leading to metabolic and signalling changes that lock the duct-like cells in their transdifferentiated state and initiate further progression to low-grade precancerous lesions.

ADM is a common and reversible process during pancreatic inflammation (pancreatitis) or injury in mouse and human tissue^4,13, and the resulting cells are believed to contribute to the regeneration of acinar structures and repopulation of the pancreas. Transgenic mouse models showed that ADM becomes irreversible when cells acquire oncogenic Kras mutations or persistent aberrant growth factor signalling, which prevent redifferentiation and initiate further progression. Oncogenic KRAS in mouse acinar cells alter gene expression profiles and lead to: silencing of acinar genes such as Mist1 (also known as Bhlha15), Cpa1 or those encoding elastase and amylase; the induction of ductal genes encoding cytokeratin 19 (Krt19) and mucin 1 (Muc1); and the upregulated expression of pancreatic and duodenal homeobox 1 (Pdx1) and Sry-related high-mobility group box 9 (Sox9)^4,11. Lineage tracing in mice has shown that acinar cells undergoing such changes due to persistent expression of oncogenic Kras^G12D transdifferentiate to ADM cells that are incapable of redifferentiating, but instead further progress to duct-like cells that form precancerous pancreatic intraepithelial neoplasia (PanIN) 1A or 1B (early dysplastic) or PanIN2 lesions (increasing levels of dysplasia)¹⁴. However, oncogenic KRAS alone does not drive carcinogenesis beyond this initiation level; secondary events are needed for further progression to carcinoma in situ (PanIN3; high-grade dysplasia) and pancreatic ductal adenocarcinoma (PDAC). Such events include additional activation of wild-type KRAS alleles through EGFR signalling^6,15–17, inflammation^5,18–20 and acquisition of additional gene mutations²¹. However, comparative studies of human tissues and transgenic mice (using the Pdx1-Cre;Kras^LSL-G12D model) suggest that ADM also can give rise to dysplastic lesions other than PanIN^22,23.

Although ADM as an initiating event for the development of pancreatic cancer has been demonstrated in mice, the proof that ADM has a role in the development of human cancer is still outstanding. On the basis of knowledge mainly obtained with genetic mouse models, this Review will discuss how acinar cell identity is maintained, how ADM (either reversible or irreversible) is initiated, as well as the currently favoured progression model via PanIN.

Acinar cell identity factors

Several basic helix–loop–helix (bHLH) transcription factors contribute to acinar cell identity and their genetic ablation in mouse models leads to dedifferentiation and ADM. Pancreas transcription factor 1 complex (PTF1) has a central role in not only maintaining the differentiation of acinar cells, but also their function by regulating the production of digestive enzymes²⁴. The PTF1 complex in adult pancreas is a trimeric transcription factor formed by recombining binding protein suppressor of hairless (RBPJ) and a dimer of the bHLH transcription factor pancreas specific transcription factor 1 alpha (PTF1A, also known as p48)²⁵. This complex then recruits p300/CREB (also known as histone acetyltransferase KAT2B), which acetylates PTF1A to further enhance transcriptional activity²⁶. This interaction can be blocked by inhibitor of β-catenin and TCF4 (ICAT; also known as β-catenin-interacting protein 1) with the net effect of negatively-regulating acinar cell differentiation²⁶. Ptf1a has been demonstrated to be epigenetically silenced during inflammation and during oncogenic KRAS-driven ADM in mice²⁷. Furthermore, ablation of Ptf1a in mice is sufficient to induce ADM, potentiate inflammation and accelerate development of invasive PDAC by sensitizing cells to KRAS-mediated transformation²⁸.

Another key regulator of proper development of the exocrine pancreas, as well as maintenance of identity and organization of adult acinar cells, is the bHLH factor MIST1, which functions as a homodimer²⁹. In acinar cells, MIST1 regulates apical–basal polarity, formation of gap junctions, proper positioning of zymogen granules and exocytosis³⁰. Acinar cells in which MIST1 homodimerization is blocked are predisposed to conversion to a duct-like phenotype, which becomes evident by increased expression of SOX9 (REF. 31), as well as upregulation of EGFR and Notch signalling pathways³². Ablation of MIST1 function leads to depletion of gap junctions, loss of polarity, dedifferentiation and ADM^31,33, but also acquisition of proliferative potential due to a decrease in p21 gene expression^34,35. In the context of Kras^G12D mice, these effects, owing to a loss of MIST1, accelerate ADM and the occurrence of PanIN³².

GATA6 is a transcription factor that maintains acinar cell differentiation by suppressing pro-inflammatory and EGFR signalling pathways. Its ablation in mice results in extensive ADM, and in the context of an activating Kras^G12V mutation, accelerates tumour development³⁶. Interestingly, smoking is a risk factor for the development of pancreatic cancer and nicotine has been shown to decrease GATA6 promoter activity, leading to loss of its expression³⁷.

Other factors that regulate acinar cell identity in mice are the bHLH transcription factor E47 (also known as TFE2), NR5A2, DICER1 and PAF1 (also known as pancreatic differentiation protein 2)^38–41. NR5A2 maintains the mature acinar differentiation state, and in the context of an oncogenic Kras mutation, loss of NR5A2 accelerates the occurrence of ADM and PanIN^40,42. Processing of microRNA (miRNA) by DICER1 is required for the maintenance of adult pancreatic acinar cells, and deletion of DICER1 increased acinar cell plasticity owing to a loss of polarity⁴¹. Additionally, deletion of DICER1 accelerates KRAS-driven acinar cell dedifferentiation and ADM, but not progression of PanIN⁴³. PAF1 expression is normally restricted to acinar cells in the pancreas, but its depletion promotes ADM, indicating a role in maintenance of acinar cell identity³⁹. Consequently, its expression is gradually lost during PDAC initiation³⁹.

Acinar cell dedifferentiation factors

CDKN1B and SOX9

Loss of cyclin-dependent kinase inhibitor 1B (CDKN1B; also known as p27^Kip1) occurs frequently in human PDAC and is associated with decreased survival⁴⁴. Nuclear CDKN1B suppresses the expression of key factors that regulate acinar cell dedifferentiation and transdifferentiation to a ductal phenotype, such as the transcription factors SOX9 and PDX1 (REF. 45). KRAS activation can decrease nuclear CDKN1B localization, which increases the expression of both SOX9 and PDX1.

In the normal (mouse and human) adult pancreas, SOX9 is expressed in centroacinar cells, at very low levels in acinar cells and in a subpopulation of ductal cells^46,47. Under inflammatory conditions, or in the presence of oncogenic KRAS, SOX9 is increasingly expressed in acinar cells and stimulates gene expression that leads to ADM⁴⁸, development of pre-malignant lesions and initiation of PDAC in mice¹⁴. In line with these findings, SOX9 expression in patient tumour samples is elevated at all stages of preneoplastic lesions and PDAC⁴⁹, correlating with increased expression of EGFR pathway-related genes⁵⁰. Similarly, in mice, the absence of SOX9 reduces EGFR signalling and pancreatic tumorigenesis⁵⁰. However, EGFR signalling can also regulate expression of SOX9 through activation of NFATC1 and NFATC4 (REFS 51,52).

PDX1

In the adult mouse pancreas, PDX1 is mainly expressed in islets and only at low levels in acinar cells^53,54. As shown by lineage tracing, during mouse development, PDX1-positive cells represent progenitors of all mature pancreatic cell types⁵⁵. PDX1 is involved in regulation of morphologic changes needed for branching morphogenesis during pancreas organogenesis^56,57, but it is also required at a later stage in development for differentiation of islet β cells⁵⁸ and the formation of acinar tissue⁵⁹. Transgenic persistent expression of PDX1 in mice leads to smaller pancreata, in which acinar cells are replaced by duct-like structures⁶⁰. In line with this finding, PDX1 is upregulated during pancreatitis, in all types of precursor lesions including PanIN, intraductal papillary mucinous neoplasms (IPMN) and mucinous cystic neoplasms (MCN), as well as in PDAC, pancreatic endocrine neoplasms and acinar cell carcinoma^53,60. PDX1 regulates ADM and the metaplastic phenotype through activation of signal transducer and activator of transcription 3 (STAT3)⁶⁰. STAT3 is a regulator of stem cell self-renewal and inflammation, and its activity is also upregulated via IL-6 (REF. 61) and KRAS–YAP1/TAZ signalling^62,63.

Notch signalling

Human PanIN samples show increased Notch activity⁶⁴. During mouse and zebrafish pancreas development, expression of Notch1 intracellular domain (NICD; activated Notch1) prevents differentiation of pancreatic acinar cells and endocrine and exocrine development, indicating that it functions to maintain the undifferentiated state of pancreatic precursor cells^65,66. In mice, Notch can be activated downstream of both EGFR–KRAS signalling and oncogenic KRAS activation to drive acinar cell dedifferentiation into a duct-like progenitor phenotype^9,64,67, but its activation is not sufficient to drive progression of preneoplastic lesions to invasive adenocarcinoma⁶⁴. Furthermore, NICD induces SOX9 expression⁶⁸, but SOX9 function is also required for maintaining Notch signalling⁶⁹, indicating a mechanism for signal amplification.

Other factors

In addition to the previously discussed molecules, MYC and KLF4 are other factors that are required to initiate the ADM process in mice^70,71. Moreover, ectopic expression of hepatocyte nuclear factor 6 (HNF6) in mouse or human acinar cells represses acinar genes and upregulates ductal genes⁴⁸.

Data also indicate that ADM might be induced by alteration of acinar cell polarity or cell–cell contacts⁷². For example, deletion of liver kinase B1 (LKB1, a regulator of energy homeostasis) in mouse pancreas (Pdx1-Cre;Kras^LSL-G12D model) leads to defective acinar cell polarity, cytoskeletal alterations and loss of tight junctions, with all in combination resulting in increased ADM⁷. Additionally, loss of NUMB, a protein that regulates integrins and cell junctions, results in dedifferentiation of acinar cells and accelerates the ADM process in mice in the presence of oncogenic KRAS⁷³. In addition to cell–matrix connections, E-cadherin-based cell–cell adhesions have important functions in maintaining the acinar cell phenotype. E-cadherin stability in epithelial cells is regulated by p120 catenin and deletion of this protein in epithelial lineages of the developing pancreas in mice leads to ADM and PanIN1A⁷⁴.

Inflammatory macrophages drive acinar cell dedifferentiation and reversible ADM

Inflammation can be a driver of acinar cell transdifferentiation and the resulting ADM cells might contribute to regeneration after pancreatitis^4,75. During pancreatic inflammation, cellular programmes downregulate factors that drive acinar cell identity such as MIST1 (REFS 76,77). Forced expression of MIST1 counteracts ADM and leads to dramatic increases in acinar cell death, organ damage and failure of pancreas repopulation⁷⁷. In caerulein-induced pancreatitis in mice, the formation of metaplastic ductal intermediates was also associated with increased Hedgehog signalling⁷⁸, which is necessary to prevent acinar cell damage and to facilitate regeneration. Ablation of macrophages in mice indicated that caerulein-driven ADM is dependent on the presence of macrophages⁴. Moreover, macrophages were shown to affect acinar cell identity in the absence or presence of an oncogenic Kras mutation^4,20.

Inflammatory macrophages initiate the ADM process via secretion of inflammatory mediators (FIG. 2), including IL-6 (REF. 61), TNF and CCL5 (also known as RANTES)⁴. IL-6 contributes to ADM through activation of JAK–STAT3 signalling⁶¹. TNF and CCL5 both activate NF-κB in acinar cells to induce expression of a multitude of genes including those that regulate the degradation of extracellular matrix and ADM, such as matrix metalloproteinase (MMP)-9 (REF. 4). Additionally, macrophage-secreted MMP7 might activate Notch signalling. Consequently, MMP inhibition in mice completely blocked caerulein-induced ADM⁴. Other transcription factors activated in acinar cells after inflammation that contribute to ADM are NFATC1 and NFATC4 (REFS 51,52).

Figure 2. Inflammatory macrophage-driven signalling leading to acinar-to-ductal metaplasia.

Inflammatory macrophages can initiate acinar-to-ductal metaplasia (ADM) through NF-κB activation as caused by secreted inflammatory cytokines such as TNF and the chemokine CCL5. Macrophage-secreted IL-6 contributes to ADM and the development of PDAC through JAK–STAT3 signalling. In addition, macrophage-secreted matrix metalloproteinases (MMPs) contribute to extracellular matrix (ECM) degradation and activate Notch signalling (NICD, Notch intracellular domain). Other transcription factors activated in acinar cells after inflammation and contributing to ADM are NFATC1 and NFATC4.

Oncogene-driven irreversible ADM

Oncogenic KRAS activates transcription factors similar to inflammatory macrophages (FIG. 3), but also facilitates persistent signalling resulting in irreversibility of the ADM process⁷⁹. Major signalling targets for activated KRAS during ADM are the RAF–MEK–ERK pathway, the phosphatidylinositol 3-kinase (PI3K)–AKT pathway and serine/threonine-protein kinase D1 (PRKD1).

Figure 3. KRAS-driven intrinsic signalling pathways leading to irreversible acinar-to-ductal metaplasia in mice.

To initiate irreversible acinar-to-ductal metaplasia (ADM), both oncogenic KRAS and wild-type KRAS activities need to be increased. This KRAS signalling mediates induction of similar transcription factors to inflammatory macrophages, but facilitates persistent signalling leading to irreversibility of the ADM process. Signalling hubs downstream of wild-type and mutant KRAS that relay signals to activate transcription factors driving ADM are PRKD1 and phosphatidylinositol 3-kinase (PI3K). PRKD1 can be activated by mutant KRAS-initiated metabolic changes and increases in mitochondrial reactive oxygen species (mROS). PRKD1 then initiates NF-κB and Notch (NICD; Notch intracellular domain) signalling and upregulated expression of matrix metalloproteinases (MMPs), epidermal growth factor (EGF), EGF receptor (EGFR) and transforming growth factor (TGF)-α (via NF-κB), and SOX9 and PDX1 (via NCID). Increased intrinsic EGFR signalling leads to further activation of wild-type KRAS and signal amplification. MMPs can contribute to extracellular matrix degradation as well as activation of Notch. PI3K induces cytoskeletal reorganization by activating small GTPases such as RAC1 and RHOA, but also activates ERK1/2 and AKT. STAT3 and NFATC1 or NFATC4 are activated via PI3K or AKT signalling. NFATC1 or NFATC4 mediate upregulation of SOX2 and SOX9. Mutant KRAS also upregulates the expression of intercellular adhesion molecule 1 (ICAM1), a surface molecule that initiates chemoattraction of inflammatory macrophages into the ADM region. Red arrows indicate signalling mediated by oncogenic KRAS. Blue arrows indicate signalling mediated by EGFR or wild-type KRAS. Grey arrows indicate signalling downstream of both mutant and wild-type KRAS. Green arrows indicate feedback signalling that potentiates KRAS-initiated signalling.

The PI3K–AKT pathway

PI3K acts downstream of KRAS and in mice, oncogenic KRAS-induced plasticity of pancreatic cells, formation of preneoplastic lesions and cancer initiation are all dependent on p110α (also known as PIK3CA, the catalytic subunit of PI3K)^80,81. In line with this finding, ADM, PanIN and the formation of invasive PDAC can also be observed after transgenic expression of a constitutively-active form of p110α⁸². PI3K-mediated transdifferentiation of acinar cells is mediated through ERK1/2 signalling⁸², and small molecule inhibitors targeting the activation of ERK1/2 indicate that these MAP kinases are involved in Kras^G12D-driven dedifferentiation of acinar cells, ADM and PanIN formation^11,83. To drive these processes, PI3K also initiates actin reorganization processes that are orchestrated by Rho GTPases^80,81,84. Pancreas-specific deletion of phosphatase and tensin homolog (PTEN), which negatively regulates PI3K signalling, leads to ductal metaplasia and malignant transformation in mice⁸⁵. In the context of an oncogenic Kras mutation, loss of PTEN leads to even more accelerated formation of PDAC^86,87. Similarly, expression of a constitutively-active allele of Akt1, one of the downstream targets for PI3K signalling, induces ADM⁸⁸ and cooperates with Kras oncogenic mutations to drive the onset and progression of PDAC⁸⁹. However, only a subset (2–3%) of human patients with pancreatic cancer carry activating mutations in PIK3CA⁹⁰, which suggests that increased PI3K activity might mainly be achieved by signalling through oncogenic KRAS in patients.

The PRKD1 pathway

Another emerging signalling pathway that drives ADM and progression to PanIN in mice (p48-Cre;Kras^LSL-G12D mouse model) is regulated by PRKD1. This enzyme converges signalling initiated by oncogenic KRAS and wild-type KRAS downstream of EGFR^6,9, and increases Notch1 activity to upregulate SOX9 and PDX1. This process is mediated through PRKD1-induced downregulation of suppressors of Notch signalling, such as SEL1L and CBL⁹, and upregulation of inducers of Notch activation, such as ADAM10, ADAM17 and MMP7 (REFS 9,91). PRKD1 also links oncogenic KRAS signalling to activation of NF-κB⁶, and activation of the PRKD1–NF-κB pathway is driven by metabolic changes initiated by KRAS that lead to an increase in mitochondrial reactive oxygen species (ROS)⁶. ROS–PRKD1–NF-κB signalling in acinar cells then upregulates expression of EGFR and its ligands, TGFα and EGF, further potentiating the oncogenic effects of mutant KRAS in a feedback loop⁶. Notch and NF-κB signalling pathways can cooperate to mediate formation of preneoplastic lesions⁹². Thus, PRKD1 brings together two important pathways that drive the formation of precancerous lesions. Notch can also act synergistically with other transcription factors such as STAT3 and the combined inhibition of Notch and JAK2–STAT3 signalling in Kras^LSL-G12D/+;Trp53^−/+;Pdx1-Cre (KPC) mice has been shown to impair ADM and its progression⁹³.

Oncogene-driven microinflammation

Acinar cells with an oncogenic Kras^G12D mutation have also been shown to produce chemoattractants for inflammatory macrophages¹⁸. This process causes a persistent microinflammation that contributes to acinar cell transdifferentiation. One of the factors released by acinar cells is intracellular adhesion molecule 1 (ICAM1, also known as CD54), which can be shed as a soluble form. Blocking ICAM1 using neutralization antibodies has been shown to substantially reduce the occurrence and progression of Kras^G12D-driven preneoplastic lesions in mice²⁰. However, such microinflammation is not sufficient to drive the progression to PDAC, and additional inflammatory insults and genetic alterations are needed for acceleration of the oncogenic process^18,19,94.

ADM as a precursor lesion

Cancer initiation

Mutations in the KRAS proto-oncogenes are the earliest events leading to development of human PDAC⁹⁵. Data from genetic mouse models have shown that transgenic expression of oncogenic KRAS in acinar cells initiates ADM and locks them into a transdifferentiated duct-like state. To progress from ADM to PanIN and pancreatic cancer, the activities of endogenous and mutant alleles of Kras need to be further increased¹⁶ (FIG. 4). Such increases in KRAS activity can be achieved by additional activation of growth factor signalling or chronic inflammation^15,17–20. EGFR signalling, for example, not only further activates oncogenic KRAS but also activates the wild-type allele^9,15,96. Upregulation of inflammatory and EGFR signalling pathways can also be achieved by loss of GATA6 (REFS 16,36,37).

Figure 4. Oncogenic KRAS and inflammation as drivers of acinar-to-ductal metaplasia and clonal expansion.

Schematic showing how macrophage subtypes and genetic mutations contribute to acinar-to-ductal metaplasia (ADM), clonal expansion and progression to pancreatic cancer. During pancreatitis ADM is a reversible process, but becomes irreversible when an oncogenic Kras mutation is present. The accumulation of KRAS activity as caused by oncogenic Kras mutations and epidermal growth factor receptor (EGFR)–wild-type KRAS signalling, as well as loss of senescence due to an additional inactivation of cyclin-dependent kinase inhibitor 2A (CDKN2A, also known as p16^INK1A), is needed for progression. Further progression to pancreatic intraepithelial neoplasia (PanIN)-2, carcinoma in situ (PanIN3) and pancreatic ductal adenocarcinoma (PDAC) occurs after acquisition of additional gene mutations inTp53 (p53), Brca2 and Smad4. The progression to cancerous lesions occurs with an increase in desmoplasia. Cells positive for the serine/threonine-protein kinase DCLK1 are of acinar origin, are formed mainly in low-grade PanIN lesions (PanIN1A, PanIN1B and PanIN2) and have cancer stem cell functions.

During the process of cancer initiation, crosstalk between acinar cells with Kras mutations and inflammatory macrophages contributes to ADM and formation of early lesions²⁰. However, during progression to PDAC, the tumour microenvironment becomes immuno-suppressive with a predominance of myeloid-derived suppressor cells and regulatory T cells⁹⁷. Furthermore, desmoplasia increases with progression⁹⁸. Pancreatic cancers can have different stromal subtypes, such as stroma characterized by stellate cell expression profiles, or more aggressive stroma characterized by activated fibroblasts and alternatively-activated macrophages⁹⁹.

Further progression

In the current model for PDAC development, ADM cells can progress to PanIN1A or PanIN1B lesions and PanIN2 lesions, which are high in senescence markers. Clonal expansion and progression to PanIN3 and PDAC requires additional signalling and mutational events to overcome oncogene-induced senescence, such as loss of cyclin-dependent kinase inhibitor 2A (CDKN2A, also known as p16^INK4A). Eventually, additional inactivating mutations of tumour suppressor genes, such as Tp53, Brca2 and Smad4, occur during PanIN2 or PanIN3 progression, but at reduced rates (a detailed review on pancreatic cancer biology and genetics and photomicrographs of different lesions can be found elsewhere¹⁰⁰). However, comparative studies of human tissues and transgenic mice (Pdx1-Cre;Kras^LSL-G12D model) suggest that dysplastic lesions other than PanIN can also arise from ADM. Such atypical flat lesions might indicate pancreatic cancer development directly from ADM without the intermediate step of PanIN^22,23.

Early dissemination and stemness

An interesting aspect during development of PDAC is that cells can disseminate from low-grade lesions with inflammatory foci, with circulating pancreatic epithelial cells present in the blood stream of mice and patients before the development of cancer^101,102. Some cells in PanIN1 or PanIN2 lesions were shown to undergo epithelial-to-mesenchymal transition (EMT)¹⁰¹, a programme that enables cells to gain invasive properties. In addition, ADM and PanIN1 or PanIN2 lesions contain a subpopulation of cells positive for the serine/threonine-protein kinase DCLK1 (also known as doublecortin-like kinase 1), and the acinar origin of these DCLK-positive cells has been demonstrated by lineage tracing¹⁰³. A majority of the circulating pancreatic epithelial cells express DCLK1 as a marker; interestingly, DCLK1 expression has also been linked to EMT¹⁰⁴.

EMT not only leads to increased invasiveness of cells, but it can also induce stem cell formation¹⁰⁵. In low-grade PanIN, DCLK1-positive cells were also shown to characterize a subpopulation with cancer stem cell properties^106,107. These cells are characterized by upregulation of Notch and EGFR signalling^103,106. However, the signalling pathways that drive the development of DCLK1-positive cells from a clonal population within ADM or PanIN cells, as well as their functions, are not yet well characterized.

Is ADM an initiating event for human PanIN and pancreatic cancer?

Although mouse data obtained with different model systems point to oncogene-driven ADM as an initiating event for the formation of PanIN lesions, the role of ADM in the development of human PDAC is still undefined. That ADM occurs in human pancreatic cancer specimens is generally accepted, as it can be observed in proximity to neoplastic precursor lesions^108,109. Attempts have been made to investigate if human acinar cells that underwent ADM can be precursors to PanIN. Analyses of human ADM lesions for KRAS mutations indicated that sections with ADM associated with PanIN lesions harboured the same KRAS gene mutation. By contrast, ADM lesions that were not associated with PanIN had wild-type KRAS. The conclusion was that the ADM lesion associated with PanIN might represent retrograde extension of the PanIN⁹⁵. With the knowledge that inflammation and macrophage-released cytokines can lead to ADM independent of KRAS mutations⁶, the detection of ADM lesions that are KRAS wild-type is not surprising. As human PDAC often has pancreatitis associated, one would expect both ADM lesions that express wild-type KRAS and ADM lesions that express mutant KRAS⁹⁵. Thus, these data can also be interpreted differently: PanIN and ADM lesions associated with PanIN have the same mutations because ADM is a precursor for PanIN; and some of the ADM have progressed to PanIN owing to additional signalling or mutations that other ADM did not have. As discussed earlier, it is also possible that the development of human PDAC from ADM might not follow the PanIN progression model, but rather might lead to occurrence of flat lesions²².

Conclusions

Although this Review focuses on acinar cell transdifferentiation as an initiating step, data does exist that supports other pancreas cell types, including duct cells or centroacinar cells, as the tumour-initiating population. For example, centroacinar cell markers were detected in patient PanINs, which led to a progression model with centroacinar cells as the origin for PDAC¹¹⁰.

In humans, PanIN are the most common of the precursor lesions for pancreatic cancer and are usually found in medium-sized ducts, whereas IPMN and MCN are found in the main duct and its major branches¹¹¹. With respect to acinar cells as potential progenitors for pancreatic lesions, accumulated conclusive evidence obtained from genetic animal models shows that mature acinar cells in the presence of an oncogenic Kras mutation transdifferentiate to the duct-like cells that form PanIN lesions¹¹². This finding was demonstrated with lineage tracing experiments^113–115, but was also shown in different animal models in which expression of oncogenic KRAS under acinar cell-specific promoters, such as PTF1A, elastase or MIST1, all induce ADM and PanIN in mice¹¹⁶. The high plasticity of acinar cells, needed for regeneration processes after pancreatic injury, also makes them vulnerable for persistent transdifferentiation to PanIN cells in the presence of an oncogenic insult.

One could argue that, with increasing age, ADM and low-grade PanIN are relatively common in humans and rarely progress to pancreatic cancer¹¹⁷. Additionally, PanIN are small, clinically difficult to detect (in contrast to IPMN) and the main focus should be on developing new treatment strategies for metastatic disease. However, unexpected findings suggest that early (low-grade) lesions produce cancer stem cells^103,107, and that epithelial cells might disseminate into the blood stream at a time point at which no primary tumour is formed in the pancreas¹⁰¹. If seeding potentially occurs at an early stage of PanIN progression, the development of an efficient treatment strategy for the time of diagnosis will be difficult, and a focus should be on detecting and intervening with these early events¹¹⁸. With respect to early detection, circulating factors that have been released by oncogenic KRAS-expressing PanIN cells might be detected in pancreatic juice or blood. The challenge in identifying reliable markers, however, is that they need to be indicative for developing cancer and distinct from factors released during pancreatitis. Eventually, understanding the crosstalk between precancerous and/or cancerous cells with cells of the desmoplastic microenvironment could be of importance to reprogramming pro-tumorigenic signalling into antitumorigenic signalling¹¹⁹.

Key points.

• Adult pancreatic acinar cells show high plasticity that enables a change in their differentiation commitment

• Acinar-to-ductal metaplasia (ADM) is a mechanism needed for regeneration after inflammation or injury

• ADM is a result of epigenetic silencing of markers of acinar cell identity, activation of drivers of acinar cell dedifferentiation or loss of acinar cell organization

• ADM is driven by intrinsic and extrinsic signalling

• ADM in the presence of oncogenic KRAS signalling is irreversible and leads to a duct-like cell type that forms pancreatic intraepithelial neoplasia