Kras and Lkb1 mutations synergistically induce intraductal papillary mucinous neoplasm derived from pancreatic duct cells
=========================================================================================================================

* Louis Collet
* Elsa Ghurburrun
* Nora Meyers
* Mohamad Assi
* Boris Pirlot
* Isabelle A Leclercq
* Anne Couvelard
* Mina Komuta
* Jérôme Cros
* Pieter Demetter
* Frédéric P Lemaigre
* Ivan Borbath
* Patrick Jacquemin

## Abstract

**Objective** Pancreatic cancer can arise from precursor lesions called intraductal papillary mucinous neoplasms (IPMN), which are characterised by cysts containing papillae and mucus-producing cells. The high frequency of *KRAS* mutations in IPMN and histological analyses suggest that oncogenic KRAS drives IPMN development from pancreatic duct cells. However, induction of *Kras* mutation in ductal cells is not sufficient to generate IPMN, and formal proof of a ductal origin of IPMN is still missing. Here we explore whether combining oncogenic *Kras* G12D mutation with an additional gene mutation known to occur in human IPMN can induce IPMN from pancreatic duct cells.

**Design** We created and phenotyped mouse models in which mutations in *Kras* and in the tumour suppressor gene liver kinase B1 (*Lkb1*/*Stk11*) are conditionally induced in pancreatic ducts using Cre-mediated gene recombination. We also tested the effect of β-catenin inhibition during formation of the lesions.

**Results** Activating *Kras* G12D mutation and *Lkb1* inactivation synergised to induce IPMN, mainly of gastric type and with malignant potential. The mouse lesions shared several features with human IPMN. Time course analysis suggested that IPMN developed from intraductal papillae and glandular neoplasms, which both derived from the epithelium lining large pancreatic ducts. β-catenin was required for the development of glandular neoplasms and subsequent development of the mucinous cells in IPMN. Instead, the lack of β-catenin did not impede formation of intraductal papillae and their progression to papillary lesions in IPMN.

**Conclusion** Our work demonstrates that IPMN can result from synergy between *Kras* G12D mutation and inactivation of a tumour suppressor gene. The ductal epithelium can give rise to glandular neoplasms and papillary lesions, which probably both contribute to IPMN formation.

*   pancreatic tumorigenesis
*   pancreatic ductal adenocarcinoma
*   IPMN
*   mucinous neoplasm

### Significance of this study

#### What is already known on this subject?

*   A significant proportion of pancreatic ductal adenocarcinoma (PDAC) cases arise from intraductal papillary mucinous neoplasm (IPMN).

*   High frequency of Kras mutations in IPMN and histological observations suggest that mutant Kras acts as a driver for IPMN development and that these IPMNs originate from pancreatic duct cells.

*   This view is challenged as Kras mutations in duct cells are not sufficient to generate IPMN.

*   Liver kinase B1 (LKB1) loss-of-function mutations are found in a significant proportion of IPMN, and they can cause Peutz-Jeghers syndrome, a disease associated with more than 100-fold increased risk to develop PDAC.

#### What are the new findings?

*   Combined activation of KRAS and inactivation of LKB1 specifically in pancreatic duct epithelium leads to formation of IPMN.

*   IPMN development resulting from KRAS and LKB1 mutations is initiated by a formation in large pancreatic ducts of intraductal papillae and glandular neoplasms, which subsequently evolve into IPMN.

*   Development of glandular neoplasms is β-catenin-dependent, whereas development of intraductal papillae is β-catenin-independent.

*   IPMN-forming lesions have a double origin and they evolve according to different molecular mechanisms.

### Significance of this study

#### How might it impact on clinical practice in the foreseeable future?

*   Drugs targeting the Wnt/β-catenin pathway could slow down IPMN progression.

*   As it is becoming increasingly clear that the neoplastic lesions (IPMN and pancreatic intraepithelial neoplasia) that cause PDAC have different cell origins and most likely evolve through different pathways, clinical practice will in the future pay greater attention to the cell type of PDAC origin.

*   Determining this origin may influence the therapeutic strategy to be implemented when confronted with precancerous lesions and PDAC.

## Introduction

Pancreatic ductal adenocarcinoma (PDAC) is one of the deadliest cancers, with a survival rate at 5 years of only 6%–7%. This results from late-stage detection of the disease, incompatibility with surgical cure and from limited progress in the development of efficient drugs. To improve patient outcome, diagnosis at an early stage is necessary, as well as a better understanding of the tumourigenic mechanisms, starting from the development of precancerous lesions.1 

PDAC is predominantly initiated by neoplastic lesions called pancreatic intraepithelial neoplasia (PanIN) and intraductal papillary mucinous neoplasm (IPMN). Lineage tracing using genetically engineered mouse models (GEMM) revealed that PanIN originates from pancreatic acinar cells.2 In adult mice, PanIN forms when two conditions are met, namely the presence of an oncogenic mutation of *Kras* (*Kras* G12D) in acinar cells, and pancreatic inflammation which in mice is usually induced by cerulein injections. Mature acinar cells then undergo acinar-to-ductal metaplasia, during which they switch to a duct-like phenotype and subsequently evolve to PanIN.3 

How IPMN is formed and how it progresses to cancer are less well known. Only a limited number of mouse models of IPMN are available,4–10 and they share the feature of initiating genetic mutations in pancreatic progenitor cells during embryogenesis. Embryonic pancreatic cells are more responsive to oncogenic stress than adult cells, and early-onset mutations are unlikely to faithfully recapitulate tumourigenesis from adult cells as it occurs in humans. Indeed, *Kras* mutations in embryonic progenitors are sufficient to generate postnatal PDAC, whereas *Kras* mutations induced in adult acini can only lead to PDAC when combined with inflammation.11 

Moreover, when mutations are induced in embryonic pancreatic progenitors, all pancreatic cell types (acinar, ductal and endocrine) are affected in their genome at birth, and identification of the cell type of IPMN origin is no longer possible. Therefore, there is a need for GEMM driving IPMN formation from well-identified postnatal pancreatic cells. Pathological analyses of human IPMN specimens suggest that IPMN originates from ductal cells, but formal proof of such origin is still missing. Genetic analyses uncovered frequent *KRAS* mutations in IPMN, with a prevalence of 50%–70%.12–14 However, since induction of a constitutively active *Kras* G12D mutation in pancreatic ducts of GEMM failed to generate IPMN, even in the presence of inflammation,2 it is unlikely that *KRAS* mutations in human pancreatic ducts are sufficient to generate IPMN.

Here we hypothesise that synergism between *Kras* G12D and another gene mutation leads to IPMN from pancreatic ducts. We considered liver kinase B1 (*Lkb1*) as a candidate since *LKB1* loss-of-function mutations are found in 4%–25% of IPMN,12 13 15 and can also cause Peutz-Jeghers syndrome, a disease associated with more than 100-fold increased risk to develop PDAC.16 17 We found that combined activation of KRAS and inactivation of LKB1 specifically in pancreatic duct epithelium leads to formation of IPMN. Since *Kras* or *Lkb1* mutations do not separately induce neoplasia,2 18 we propose that a combination of mutations is required for the development of IPMN from ductal cells.

## Materials and methods

### Immunofluorescence and immunohistochemistry

Dissected pancreata were fixed in 4% paraformaldehyde at 4°C for 6 hours before embedding in paraffin. Immunofluorescence and immunohistochemistry (IHC) were performed on 6 µm tissue sections according to the protocol described in Prévot *et al*.19 Primary antibodies are listed in online supplementary table 1.

### Supplementary data

[[gutjnl-2018-318059supp001.pdf]](pending:yes) 

For IHC, antibody binding was visualised by a biotinylated secondary antibody (1/1000), a streptavidin-POD conjugate (1/1000) (Sigma-Aldrich, Bornem, Belgium) and 3,3’-diaminobenzidine tetrachloride (Abcam, Cambridge, UK) as a substrate, and haematoxylin was used to counterstain the tissue. Slides stained by IHC were scanned by Mirax Imaging system and the Mirax Viewer (Zeiss, Zaventem, Belgium) software was used to capture images.

For immunofluorescence labelling, secondary antibodies were applied at 1/1000 dilution and nuclei were labelled by DAPI (4’,6'-diamidino-2-phenylindole). Photographs were taken by Axiovert 200 fluorescent microscope (Zeiss) using the AxioVision program.

To detect fibrosis, slides were incubated into a picric acid solution with Sirius Red (Direct Red 80, Sigma-Aldrich) and Fast Green (Sigma-Aldrich) for 4 hours. Neoplastic lesions were detected by Alcian blue staining with an eosin counterstaining.

### Quantification procedures and statistical methods

The size and area of ductal hyperplasia were analysed using the Adobe Photoshop CS5 Extended program. Duct size was determined by measuring the longest width of the normal and dilated ducts from H&E pictures (JPG files). Statistical analysis was performed using a two-tailed unpaired t-test (without Welch’s correction).

To determine the percentage of transformed area in KLC pancreas at 10 weeks, we used the CaseViewer software (3DHistech, Budapest, Hungary). The percentage of transformed area is equal to the sum of transformed surfaces divided by the whole surface of pancreatic parenchyma. One-sample two-tailed t-test was used to perform the statistical analysis.

The percentage of Alcian blue-positive surfaces relative to the total transformed area was calculated using the Area Quantification V.1.0 plugin of the Halo software (Indica Labs, Corrales, USA). Two-tailed unpaired t-test with Welch’s correction (which does not assume equal variance) was performed to compare means using the Prism V.5.0a software (GraphPad, San Diego, USA).

Additional detailed descriptions of all materials and experimental methods can be found in online supplementary methods.

## Results

### Ductal Lkb1 and KrasG12D mutations synergise to generate gastric-type IPMN

To investigate the potential development of IPMN from ductal cells, we combined *Kras* G12D mutation with inactivation of LKB1, a tumour suppressor known to be mutated in human IPMN.12 13 15 Control (Ctrl), LSL-KrasG12D/Sox9CreER (KC), LKB1f/f/Sox9CreER (LC) and LSL-KrasG12D/LKB1f/f/Sox9CreER (KLC) mice were treated at the adult stage with tamoxifen (see online supplementary figure 1A). Eight weeks later, KLC mice showed about 20% weight loss and were sacrificed for ethical reasons. Gross examination of their pancreas revealed the presence of a cystic pancreas (data not shown), and histological examination showed multiple cysts intermingled with papillary lesions. LC and KC mice did not develop detectable lesions 8 weeks after tamoxifen injection (see online supplementary figure 1B; data not shown).

### Supplementary data

[[gutjnl-2018-318059supp002.pdf]](pending:yes) 

Newborn KLC mice were also treated with tamoxifen in parallel experiments and developed lesions identical to those seen in adult mice. Since performing tamoxifen injections in newborn mice is time-saving, we pursued our work with these mice. Ten weeks after tamoxifen administration (figure 1A), KLC mice showed weight loss and were sacrificed. Cysts were macroscopically visible on KLC pancreata (figure 1B). In all the mice analysed (n=8), papillary lesions and cysts, which were located both in the head and tail of the pancreata, were detected following H&E staining; Sirius Red staining revealed extensive fibrosis (figure 1C). Large parts of pancreata stained positively for Alcian blue. Of the KLC pancreata, 99.61% appeared transformed (SD=0.47%, n=3, p<0.0001), while no lesion was detected in KC mice (data not shown); LC mice displayed areas with ductal dilations and lipomatosis (figure 1C).18 These results indicate that KRAS activation and LKB1 inactivation synergise to generate dysplastic lesions from the ductal compartment.

![Figure 1](http://gut.bmj.com/https://gut.bmj.com/content/gutjnl/69/4/704/F1.medium.gif)

[Figure 1](http://gut.bmj.com/content/69/4/704/F1)

Figure 1 
IPMNs are detected in KLC pancreas. (A) Timing of tamoxifen injections to induce *Kras* and *Lkb1* mutations in newborn mice. (B) Macroscopic view of a 10-week-old KLC mouse showing a pancreas with small cysts and a mucinous appearance. (C) Histological stainings on Ctrl, LC and KLC pancreata. Ductal hyperplasia (black arrow) and lipomatosis (red arrow) are present in LC pancreas, and cysts (C) and papillary lesions (arrowheads) are detected in KLC pancreas. KC pancreas shows no phenotype (not shown). (D) Morphometric analysis of pancreata from 10-week-old LC (n=3) and KLC (n=2) mice. The graph shows the number of ductal lesions with the greatest diameter >400 µm (0±0/field vs 2.35±0.37/field, mean±SEM; \***|P<0.001). No ductal dilation larger than 400 µm is observed in Ctrl or LC pancreas (not shown). (E) Small mucinous lesions are seen in KLC pancreas at 10 weeks. (F) High-grade dysplasia with increased cell density, nuclear pseudostratifications and irregular nuclei are observed in IPMN from KLC mice. Bar=200 µm. Ctrl, control; IPMN, intraductal papillary mucinous neoplasm; KLC, LSL-KrasG12D/LKB1f/f/Sox9CreER; LC, LKB1f/f/Sox9CreER.

IPMN is defined in mice by the presence of large papillae with mucinous cells and of cysts larger than 400 µm (figure 1C).4 The size of papillary lesions and cysts in KLC pancreata fitted with the histological definition of IPMN (figure 1D). Lineage tracing using KLC/ROSA26eYFP mice revealed that IPMN derived from the ductal compartment (see online supplementary figure 1C). Next to the large papillary lesions described above, smaller mucinous lesions were also found in KLC pancreata (figure 1E). These lesions were YFP-positive, indicating that they were formed from the ductal compartment (see online supplementary figure 1D). They are likely early-stage lesions which will evolve to larger lesions. Low-grade dysplasia was predominant in IPMN lesions, although high-grade dysplasia was also observed (figure 1F and see online supplementary figure 2 and table 2).

Many IPMN cells in KLC pancreata have a columnar shape, with basal nuclei and cytoplasmic mucin (see online supplementary figure 1E). This suggests that lesions are of gastric type. To confirm this, we performed immunostaining to detect mucin (Muc) 1, 2 and 5AC, and CDX2. While IPMN lesions from KLC pancreas were Muc1-positive and Muc5AC-positive, they were negative for Muc2 and CDX2 (see online supplementary figure 3A–D). Furthermore, the stroma of KLC pancreata showed no to very little ERα expression (see online supplementary figure 3E), indicating that these lesions could not be considered as mucinous cystic neoplasm. Overall, these data indicate that KLC pancreata display key characteristics of gastric-type IPMN (see also online supplementary table 2).

To determine whether the combination of *Kras* G12D and *Lkb1* mutations was also able to induce lesions from acinar cells, we generated LSL-KrasG12D/LKB1f/f/ElaCreER (KLE) mice. Similar to KLC mice, KLE mice were treated at birth with tamoxifen and sacrificed at 10 weeks of age. Their pancreata did not show obvious lesions (see online supplementary figure 5), indicating that duct cells, but not acinar cells, develop into dysplastic lesions when exposed to *Kras* G12D and *Lkb1* mutations.

### Glandular neoplasms develop near large pancreatic ducts in KLC mice

To understand how pancreatic lesions are initiated in KLC mice, we collected pancreata 2 and 4 weeks after tamoxifen treatment (figure 1A). Careful observation of KLC pancreas sections 2 weeks after tamoxifen treatment (n=3) revealed the presence of small evaginating buds associated with large ducts in KLC pancreata (figure 2A,B). Developing intraductal papillae were also detected at that stage (figure 2B). At 4 weeks (n=3), KLC pancreata were normal except that the evaginating buds have evolved to glandular neoplasms whose lumen communicated directly with the nearby pancreatic duct (figure 2C and online supplementary figure 6A). Glandular neoplasms were only found in KLC pancreata, and not in Ctrl or LC pancreata (figure 2C). Their epithelium often stained positive for Alcian blue (figure 2D). The ducts retained an otherwise normal histology (figure 2A,C,D), although the presence of intraductal papillae could also be noted (see online supplementary figure 6A, arrows). Quantification of the two duct-associated lesions revealed that glandular neoplasms were 5.8 times more abundant than intraductal papillae at 4 weeks. The well-preserved morphology of KLC pancreata at 4 weeks also showed that neoplasms were found only in the vicinity of the main pancreatic duct and interlobular ducts draining secretions from the pancreatic head and tail (defined here as large ducts); these neoplasms were not present near intralobular or intercalated ducts, or centroacinar cells (see online supplementary figures 6B and 7).

![Figure 2](http://gut.bmj.com/https://gut.bmj.com/content/gutjnl/69/4/704/F2.medium.gif)

[Figure 2](http://gut.bmj.com/content/69/4/704/F2)

Figure 2 
Early neoplastic lesions with active ERK are closely associated with large pancreatic ducts in KLC mice. (A) At 2 weeks, small dilations of the ductal epithelium are observed in KLC pancreas (arrows). Such structures are not detected in Ctrl, LC and KC pancreas (not shown). Bar=20 µm. (B) Sox9 and E-cadherin colabelling confirms the presence of small dilations associated with large ducts in KLC pancreas. Papillary projections from the epithelium are also detectable (arrow). Early-stage structures are observed in 58.8%±2.5% of the large ducts in KLC pancreata (mean±SEM; n=3 mice). Such structures emerging from large ducts are not seen in Ctrl, LC and KC pancreata (not shown). i, islet. Bar=50 µm. (C) At 4 weeks, large ducts appear normal in Ctrl and LC mice. The number of large duct sections examined in Ctrl and LC mice is 36 and 42, respectively (number of mice examined, n=3 in both cases). In contrast, KLC mice exhibit multiple neoplastic lesions located near their large ducts. The glandular neoplasms are connected to the lumen of large ducts in KLC pancreas. Quantification of the percentage of large ducts having lesions in KLC pancreata indicated that glandular and papillary lesions were detected in 86.5%±4.4% and 41.3%±10.7% of these ducts, respectively (mean±SEM; n=3 mice; p=0.003 for glandular lesions and p=0.061 for papillary lesions). (D) Glandular neoplasms near large ducts of KLC pancreas stain positively for Alcian blue. Bar=200 µm. (E) P-ERK immunostaining of Ctrl, LC and KLC pancreata at 2, 4 and 10 weeks. At 2 weeks, the ERK pathway is active in large ducts of KLC pancreas and the associated dilations (black arrow), whereas no or weak activity is observed in the large ducts of Ctrl and LC mice. ERK activity is also observed in neoplasms, cysts and IPMN of KLC pancreata at 4 and 10 weeks. At 4 weeks, papillae positive for P-ERK (red arrows) are sometimes observed in the lumen of a large KLC duct. At 4 and 10 weeks, P-ERK staining stays low in Ctrl and KC pancreata (not shown). i, islet of Langerhans. Bar=50 µm. Ctrl, control; DAPI, 4’,6'-diamidino-2-phenylindole; KC, LSL-KrasG12D/Sox9CreER; KLC, LSL-KrasG12D/LKB1f/f/Sox9CreER; LC, LKB1f/f/Sox9CreER.

While our data provide evidence that the glandular neoplasms in KLC mice derive from evaginating ductal epithelium, their localisation may suggest that they derive from pancreatic duct glands (PDG) which have been described near large pancreatic ducts.10 20 21 However, we could not detect PDG associated with large pancreatic ducts in Ctrl and LC pancreata (figure 2A,C,D), and considered highly unlikely that PDG might be at the origin of glandular neoplasms.

We concluded that evaginating buds and small intraductal papillae observed at 2 weeks evolve towards the larger glandular neoplasms and intraductal papillae detected in 4-week-old KLC pancreas and eventually into IPMN at 10 weeks.

### Early activation of the MAPK-ERK pathway confirms that IPMN originates from cells located in the large pancreatic ducts of KLC mice

Oncogenic KRAS typically induces activation of the ERK and AKT pathways. To test whether these pathways were activated in our models, we performed P-ERK and P-AKT immunostaining on pancreata of 2-week-old, 4-week-old and 10-week-old mice. Ctrl and LC mice showed low ERK activity in the large ducts 2 weeks after tamoxifen injections (figure 2E). By contrast, in KLC mice, higher P-ERK was detected in large ducts, especially at the small buds evaginating from the large ducts (figure 2E, black arrow). Four weeks after treatment, ERK phosphorylation was strongly increased in the glandular neoplasms associated with the large ducts of KLC mice. In addition, intraductal papillae were also positive for P-ERK (figure 2E, red arrows). At 10 weeks, IPMN present in KLC pancreata showed strong P-ERK staining (figure 2E and online supplementary figure 4D). Thus, it appears that ERK signalling is active as soon as the earliest lesions are formed, both in IPMN and PanIN (see online supplementary figure 8A).

AKT phosphorylation was absent in the large ducts of the different genotypes at 2 weeks. However, at 4 weeks, P-AKT staining was increased in large ducts of KLC pancreata, as well as Ctrl and LC pancreata (see online supplementary figure 9A). At 10 weeks, AKT phosphorylation was observed in large ducts of Ctrl and LC mice, and in IPMN of KLC mice, respectively (see online supplementary figure 9A). The detection of phosphorylated GSK3β, a downstream target of AKT, confirmed the activation of the AKT pathway in KLC pancreata (see online supplementary figure 9B). P-GSK3β and P-AKT stainings were decreased in some mucinous cells of glandular neoplasms (see online supplementary figure 9B, arrows; not shown).

Finally, since the SRC pathway is activated in pancreatic tumourigenesis, we compared SRC phosphorylation in IPMN of KLC mice and in acinar-derived PanIN.22 We found significant P-SRC immunostaining in PanIN but not in IPMN, except for typical tuft cells (previously known to be positive for P-Src) (see online supplementary figure 9C).23 

### Early modifications in ductal cell identity and macrophage infiltration are associated with IPMN development

To analyse whether IPMN development was associated with cell fate reprogramming, we characterised the expression of the ductal transcription factors Sox9, HNF6, Prox1 and HNF1β in KLC pancreas. Expression of Sox9 was maintained in glandular neoplasms and IPMN (figure 3A and online supplementary figure 1C). In contrast, expression of the ductal transcription factor HNF6 was maintained in smaller ducts, but repressed in glandular neoplasms and large pancreatic ducts (figure 3A, white and yellow arrows). This revealed that cells lining large ducts are affected by the combined *Kras* G12D and *Lkb1* mutations even if they displayed normal histology. HNF6, as well as Prox1 and HNF1β, were absent from IPMN (see online supplementary figure 10; data not shown).

![Figure 3](http://gut.bmj.com/https://gut.bmj.com/content/gutjnl/69/4/704/F3.medium.gif)

[Figure 3](http://gut.bmj.com/content/69/4/704/F3)

Figure 3 
Expression of transcription factors is modified during IPMN development. (A) Sox9 expression is maintained in neoplastic lesions near large ducts (yellow arrows). In contrast HNF6 expression disappears in these lesions and in large ducts but persisted in smaller ducts (white arrows). H&E staining of this tissue section allows better visualisation of neoplastic lesions. The bottom panel shows magnifications of the corresponding top panel. (B) The transcription factor KLF4 is neither expressed in large ducts of KLC mice at 2 and 4 weeks nor in IPMN at 10 weeks, with the exception of some mucinous cells (arrows). (C) KLF5 is expressed in large ducts of KLC pancreas at 2 weeks. KLF5 was also detected in glandular neoplasms at 4 weeks and in IPMN at 10 weeks; it disappears in some mucinous cells localised in neoplastic lesions or IPMN (black arrows). Bar=50 µm. IPMN, intraductal papillary mucinous neoplasm; KLC, LSL-KrasG12D/LKB1f/f/Sox9CreER. 

The transcription factors KLF4 and KLF5 play a role in PanIN development,24 25 but their expression in IPMN is unknown. We observed that KLF4 was expressed neither in ductal cells of Ctrl or LC pancreas at any stage nor in the large ducts of KLC pancreata at 2 weeks (not shown; figure 3B). However, KLF4 expression started to be detectable in mucinous cells of IPMN developing in KLC pancreata at 4 weeks (figure 3B, arrows). In Ctrl pancreata, KLF5 was expressed in large ducts but absent in smaller ducts (figure 3C; not shown). Although present in most neoplastic lesions and IPMN at 4 and 10 weeks, KLF5 expression was absent from mucinous cells of KLC pancreata (figure 3C, arrow). Thus the expression pattern of KLF4 appears more restricted in IPMN than in PanIN (see online supplementary figure 8B), while KLF5 is widely expressed both in IPMN and PanIN (see online supplementary figure 8C).

We concluded that expression of several ductal transcription factors (HNF6, Prox1 and HNF1β) is repressed during IPMN formation, indicating that this process is associated with partial loss of the ductal cell identity. Conversely, the expression of Sox9 and KLF5 transcription factors is maintained during this process, suggesting a role for these factors in IPMN development. Finally, the expression of KLF4 in some mucinous cells suggests the presence of a subpopulation of more differentiated IPMN cells.

To verify whether immune cells were associated with IPMN, we detected B and T lymphocytes, as well as macrophages, by immunostaining. B lymphocytes were not detected, regardless of the genotype or stage considered (data not shown). Rare CD3-positive T lymphocytes were observed in the stroma of Ctrl, LC and KLC pancreata 4 weeks after tamoxifen treatment (see online supplementary figure 11A, arrows). Very few F4/80-positive macrophages were observed in the stroma surrounding the large ducts of Ctrl and LC pancreas, at all tested stages. In contrast, many macrophages were associated with neoplastic lesions in KLC pancreata starting 4 weeks after tamoxifen treatment (see online supplementary figure 11B). Therefore, the scarcity of lymphocytes and abundance of macrophages in developing IPMN of KLC pancreata contrast with the abundance of both T cells and macrophages in acinar-derived PanIN (see online supplementary figure 11C).

### Cells from KLC pancreas grafted in immunodeficient mice have malignant potential

Premature death of KLC mice likely results from exocrine deficiency and/or of lesions potentially affecting other organs in which the Cre recombinase is active. Therefore, to investigate the malignant potential of IPMN in KLC mice, pancreata of four KLC/RosaeYFP mice were dissected, separately digested into a cell suspension, and subsequently injected into the tail of the pancreas of nude mice (n=8).

Macroscopic analysis performed 10 weeks after injection revealed the presence of tumours in the tail of the pancreas (in 8/8 mice), as well as tumours adhering to the abdominal muscles near the pancreas and in the spleen (in 8/8 mice; figure 4A). The presence of these tumours at these last two locations likely results from leakage of grafted tumour cells through the hole generated by the injection needle. Histological analysis of these tissues confirmed the presence of Alcian blue-positive tumour cells invading the pancreas, muscle tissue and spleen parenchyma (figure 4B). The tumour cells expressed YFP, Sox9 and KLF5, which were detected by immunofluorescence and IHC (figure 4C), as well as P-ERK, P-AKT, KLF4 and KLF5, which were detected by western blotting (figure 4D). These results reinforced the fact that these tumours more than likely originated from ductal cells of KLC pancreas. Furthermore, pan-cytokeratin immunolabelling provided evidence that some of these cells invade the stroma around the tumour epithelial structures (figure 4C).

![Figure 4](http://gut.bmj.com/https://gut.bmj.com/content/gutjnl/69/4/704/F4.medium.gif)

[Figure 4](http://gut.bmj.com/content/69/4/704/F4)

Figure 4 
Grafted IPMN gives rise to malignant tumours. (A) The tumours are structurally similar to IPMNs, with cysts and papillary formations. The insets show magnifications of the tumours infiltrating into the pancreas, spleen and muscle. S, spleen. Bar=200 µm. (B) Alcian blue staining highlights the presence of mucinous cells in the tumour located in the pancreas, or infiltrating the spleen and muscle. Bar=200 µm. (C) Grafted tumours express pan-cytokeratin, Sox9 and KLF5. Pan-cytokeratin labelling suggests that cells delaminate from the epithelial tumour sheet to invade the adjacent stroma (white arrows). Similar to the primary lesions, some mucinous cells in the lesions derived from the graft are negative for KLF5 (black arrows). Bar=20 µm. YFP immunolabelling of a tumour located in the pancreas of a nude mouse confirms that tumour cells originated from a KLC/RosaeYFP pancreas. Bar=50 µm. (D) Western blotting experiments performed on protein extracts from three different tumours and three Ctrl mouse pancreata with P-ERK, P-AKT, KLF4 and KLF5 antibodies show that these factors and the activity of the ERK and AKT pathways are strongly increased in grafted pancreata. Ctrl, control; DAPI, 4’,6'-diamidino-2-phenylindole; IPMN, intraductal papillary mucinous neoplasm. 

Nude mice orthotopically grafted have a strongly distended abdomen, and the morbidity generated by this situation imposes the euthanasia of the mice 10 weeks after the graft. To study the metastatic behaviour of the KLC pancreatic cells, the tail vein of five nude mice was thereby injected with a cell suspension obtained from digested KLC pancreas. Six to seven weeks later, these mice showed weight loss and were euthanised. Their liver appeared macroscopically and histologically normal (data not shown). However, the lungs of the five nude mice were infiltrated by many metastatic nodules (see online supplementary figure 12A). Histologically, these nodules measured up to 1–2 mm in diameter, and formed branched and duct-like glandular structures, with angular contours. Stroma was scant. Neoplastic cells were columnar to cuboidal and produced mucins, and their nuclei were round to ovoid, with little pleomorphism (see online supplementary figure 12A,B). Tumour cells showed variable staining with Alcian blue staining (see online supplementary figure 12A,B). Cells weakly labelled with Alcian blue strongly expressed Sox9 and KLF5, whereas strongly Alcian blue-positive cells had lower expression of Sox9 or loss of KLF5 expression (see online supplementary figure 12B), reproducing observations made in KLC pancreas. All together, these observations show that *Kras* and *Lkb1* mutations in cells of the ductal compartment can confer a malignant potential.

### β-catenin is essential for the formation of the neoplastic mucinous glandular lesions

Activating mutations in the β-catenin (CTNNB1) gene and mutations in genes resulting in increased activity of the Wnt/β-catenin pathway have been discovered in IPMN,11 suggesting a role for the Wnt/β-catenin pathway in IPMN development. To verify this hypothesis, we first tested whether an increased activity of this pathway was observed in lesions found in KLC pancreas by performing β-catenin immunolabelling. In ducts of Ctrl pancreata, β-catenin was mainly localised to the lateral membrane of the cells (figure 5A). In contrast, β-catenin accumulation was seen in the cytoplasm and nucleus of many cells in the glandular lesions of KLC pancreas at 4 weeks (figure 5A). Similar cytoplasmic and nuclear accumulation of β-catenin was also detected in IPMN found in KLC pancreas at 10 weeks (see online supplementary figure 13A). This accumulation was associated with the induction of the expression of cyclin D1, a well-known target of the Wnt/β-catenin pathway, in the glandular lesions of KLC pancreata (see online supplementary figure 13B).

![Figure 5](http://gut.bmj.com/https://gut.bmj.com/content/gutjnl/69/4/704/F5.medium.gif)

[Figure 5](http://gut.bmj.com/content/69/4/704/F5)

Figure 5 
The β-catenin pathway plays a key role in the development of glandular mucinous neoplasia. (A) β-catenin immunolabelling performed on pancreas sections of Ctrl and KLC pancreata at 4 weeks. DAPI was used to label the nuclei. The white dotted line locates the ductal epithelium in the Ctrl pancreas. The white arrow shows a cell with an absence of β-catenin in the nucleus. The yellow dotted line illustrates a group of cells showing cytoplasmic and nuclear accumulation of β-catenin in the KLC pancreas. A, acinar tissue. Bar=20 µm. (B) Four weeks after tamoxifen treatment, no glandular structure is observed in the stroma surrounding the large ducts of KLCC pancreas (H&E staining). (C) The lack of Alcian blue staining around the large ducts of KLCC pancreas confirms the absence of glandular lesions. Alcian blue staining of intestinal cells on the same section served as positive control (not shown). (D) Intraductal papillae (arrows) are seen in large ducts of KLCC pancreas. Bar=100 µm. (E) Twelve weeks after tamoxifen treatment, cystic and papillary lesions are present in KLCC pancreas. (F) Alcian blue staining indicates that most of these lesions are non-mucinous. Bar=200 µm. (G) Rare Alcian blue-positive lesions detected in KLCC pancreas stain positively for β-catenin, indicating that in these lesions β-catenin inactivation has not occurred. The picture also shows that Alcian blue-negative lesions are also β-catenin-negative. Bar=100 µm. (H) KLF5 is widely expressed in KLCC pancreata. (I) ERK activity is detected in the lesions of KLCC pancreas, but is heterogeneous: some lesions show high ERK activity (star), whereas other lesions are weakly, or not, stained (arrow). (J) Cells from KLCC lesions also express Sox9. Bar=100 µm. Ctrl, control; DAPI, 4’,6'-diamidino-2-phenylindole; KLC, LSL-KrasG12D/LKB1f/f/Sox9CreER; KLCC, LSL-KrasG12D/LKB1f/f/CTNNB1f/f/Sox9CreER.

With this result in hand, we generated LSL-KrasG12D/LKB1f/f/CTNNB1f/f/Sox9CreER (KLCC) mice. Newborn KLCC mice were treated with tamoxifen and sacrificed 4 and 12 weeks after tamoxifen administration. At 4 weeks, inflammation was found in the head of the pancreas, but the body and the tail of the pancreas were normal. Intraductal papillae were readily detected (figure 5A–C), as in KLC mice (compare figure 5C with online supplementary figure 6). However, KLCC pancreas lacked glandular neoplasms appended to large ducts (compare figure 5A with figure 2C); this was confirmed by the absence of Alcian blue-positive lesions near the large ducts (figure 5B).

At 12 weeks, KLCC pancreata were cystic and displayed an inflammatory infiltrate, like KLC pancreas (figure 5D). However, the neoplastic lesions were not mucinous as evidenced by the absence of Alcian blue staining (figure 5E), except for a few Alcian blue-positive (figure 5F) or PAS-positive (data not shown) lesions which had escaped β-catenin inactivation and still expressed β-catenin. Quantification of the surface of Alcian blue-positive lesions in KLC and KLCC mice revealed a nearly 100-fold reduction in Alcian blue lesions in KLCC mice (see online supplementary figure 13C). Most lesions were positive for KLF5 (figure 5H), P-ERK staining (figure 5I) and Sox9 (figure 5J).

We conclude that following induction of *Lkb1* and *Kras* mutations, β-catenin is required for the development of glandular neoplasms appended to the large ducts, but not for intraductal papillary lesions. Our data also suggest that mucinous lesions induced by *Lkb1* and *Kras* mutations derive mainly from these glandular neoplasms, whereas intraductal papillary neoplasia in large ducts gives rise to non-mucinous lesions.

### Human and mouse IPMN share many characteristics

Finally, we investigated whether morphological and molecular characteristics of mouse IPMN were detectable in human IPMN. In patient samples with IPMN, the surrounding pancreas showed neoplastic lesions near large ducts, like in mice (figure 6A,B). Papillary structures in human and mouse IPMN were comparable (figure 6C).

![Figure 6](http://gut.bmj.com/https://gut.bmj.com/content/gutjnl/69/4/704/F6.medium.gif)

[Figure 6](http://gut.bmj.com/content/69/4/704/F6)

Figure 6 
Human and mouse IPMNs are similar. (A) Human IPMN and KLC pancreas contain neoplastic lesions (arrows in the human panel) located near large ducts. (B) Neoplastic papillary lesions with a glandular appearance (dotted line) can be detected close to human large ducts. Bar=200 µm. (C) Papillary lesions are also similar in humans and mice. Bar=50 µm. (D) Comparison of the expression of P-ERK, KLF5 and KLF4 between human IPMN and early-stage PDAC. ERK pathway is active both in human IPMN and PanIN. Both also express KLF5, except for a small proportion of mucinous cells (arrows). Conversely, KLF4 is not detected in human IPMN with the exception of some mucinous cells (inset), while KLF4 is widely expressed in PanIN. Bar=50 µm. Star: lumen of human large ducts. Thirteen different human samples were used in this comparative analysis. IPMN, intraductal papillary mucinous neoplasm; KLC, LSL-KrasG12D/LKB1f/f/Sox9CreER; PDAC, pancreatic ductal adenocarcinoma; PanIN, pancreatic intraepithelial neoplasia.  

We also observed ERK phosphorylation in human IPMN, like in mice (figure 6D). Again like in mice, KLF5 expression was widely expressed in human IPMN, except for some mucinous cells (figure 6D). KLF4 expression was not present in human IPMN (figure 6D), with the exception of a small proportion of highly mucinous cells (inset in figure 6D), as observed in mice. A similar situation was observed in PanIN of early-stage PDAC: P-ERK and KLF5 were widely expressed (figure 6D). Nevertheless, KLF4 showed a broader expression pattern in PanIN (as previously described, ref 24) than in IPMN (figure 6D). P-ERK and KLF5 were detected in the main duct and branch duct IPMN, both in the gastric and intestinal types, while KLF4 was undetected (see online supplementary figure 14). Together, these results confirm that KLC mice are a valid model for human IPMN.

## Discussion

In the present study, we provide evidence supported by genetic lineage tracing that IPMN originates from ductal cells. To initiate IPMN development from ductal cells, we need to combine *Kras* and *Lkb1* mutations as *Kras* or *Lkb1* mutation alone is unable to generate neoplastic lesions. Our results also suggest that the development of IPMN from ductal cells and the development of PanIN from acinar cells take place via different mechanisms, as combining *Kras* and *Lkb1* mutations in acinar cells does not generate PanIN. Moreover T cells, which are important for PanIN development,26 are almost absent from IPMN pancreas. Finally, with the exception of tuft cells, SRC phosphorylation is not detected in IPMN cells, unlike in PanIN.

The first visible manifestation of IPMN in our mouse model is the formation of small buds evaginating from the ductal epithelium into the stroma surrounding large pancreatic ducts. Histological analyses provide evidence that these buds give rise to the glandular neoplasms. Intraductal papillae are also observed within the large ducts. Therefore, glandular neoplasm and intraductal papillae originate from ductal epithelial cells and probably combine to develop IPMN. How ductal cells become fated to either glandular neoplasms or papillary lesions remains unknown and may reflect heterogeneity within the ductal epithelium. Interestingly, formation of glandular neoplasms and subsequent development of mucinous cells in the IPMN are dependent on the β-catenin pathway, unlike the formation of intraductal papillae and of non-mucinous papillary projections within the IPMN (see online supplementary figure 15). This supports the notion that subpopulations of ductal cells may differentially respond to mutational injuries. In favour of this hypothesis, recent results indicate that combined *Kras* and *p53* mutations in ductal cells contribute to the formation of non-mucinous intraductal papillary lesions27; importantly, in this model, the pancreatic ducts did not form glandular neoplasms. Similarly, loss of PTEN and activation of KRAS synergistically induce formation of neoplasia from ductal cells, without development of mucinous glandular neoplasms.28 

Our study highlights the role of the β-catenin pathway in the development of IPMN. The β-catenin gene (*CTNNB1*) is mutated at a low frequency in patients with IPMN. However, mutations in the *GNAS* and *RNF43*, which encode the G-protein alpha stimulatory subunit (Gαs) of heterotrimeric G proteins and an E3 ubiquitin ligase, respectively, are the second and third most frequent mutations in patients with IPMN.29 30 Both genes are implicated in the activity of the Wnt/β-catenin pathway,31 32 suggesting that the Wnt/β-catenin pathway is perturbed in a significant proportion of IPMN.

Intriguingly, our data suggest that synergism between *Kras* and *Lkb1* mutations occurs mainly in large ducts, although we cannot formally exclude that lesions also originate from smaller ducts or centroacinar cells. To explain why such synergism would not be observed in cells lining the intralobular ducts or in centroacinar cells, a mechanism controlling the activity of KRAS in Ctrl pancreas must be considered. This mechanism, which would be deficient in the absence of LKB1, could involve translational control of KRAS, the expression of a KRAS repressor or the confinement of KRAS in a cell compartment where it cannot exert its activity. In humans, IPMNs originating from the main duct are mainly of the intestinal type. In contrast, we find that in the KLC model, the IPMNs are of gastric type. The reason for the discrepancy is not clear, but we hypothesise that the mutational profile of IPMN determines the histological subtype. In this context, it is possible that the development of distinct IPMN subtypes depends on the cell of origin and relies on distinct molecular mechanisms. Also, it would be interesting to investigate whether LKB1 mutations are associated with main duct IPMN of the intestinal or gastric type.

Neoplastic transformation is usually accompanied by cell fate reprogramming. During PanIN formation, this initially results in acinar-to-ductal metaplasia. During IPMN development, expression of ductal transcription factors (HNF6, HNF1β, Prox1) is downregulated. This is likely important for IPMN progression. Given the known proproliferative effects of KLF5,33 this factor might be important for proliferation of IPMN. Additional observations show that KLF5 is no longer expressed in some mucinous cells found in IPMN. The latter could correspond to mucinous cells that express KLF4, a factor previously involved in terminal differentiation of normal cells34 and cell cycle arrest.35 The IPMN-associated mucinous cells could be considered as differentiated cancer cells. Understanding how IPMN generates such cells is useful for therapy when considering that promoting the differentiation of cancer cells helps in favouring proliferation arrest.

In conclusion, we provide evidence that IPMN originates from ductal cells. Our results also suggest that subsets of ductal cells line large pancreatic ducts, with a subset able to form mucinous lesions and another subset which is prone to generate intraductal papillary lesions. Our findings open perspectives for improved diagnosis and better understanding of the role of oncogenic and tumour suppressor mutations in pancreas tumourigenesis.

## Acknowledgments

The authors thank the members of the LPAD laboratory and Mark Rider for help and discussions; Mourad El Kaddouri, Vitaline De Greef, Katarzyna Konobrocka and Jean-Nicolas Lodewyckx for expert technical assistance; Maike Sander and Janel Kopp for sharing unpublished observations; and Maike Sander, Doris Stoffers and Kei Sakamoto for providing mouse strains.

## Footnotes

*   LC and EG contributed equally.

*   Contributors Acquisition and statistical analysis of data: LC, EG. Analysis and interpretation of data: LC, EG, AC, IB, PJ. Provision of material: MK, PD. Drafting the manuscript: PJ, LC, EG. Revision of the manuscript: LC, EG, AC, PD, IB, FL, PJ.

*   Funding This work was supported by grants from the Fondation contre le Cancer (Belgium #2014-125 and #2018-078 to FL, and #2016-089 and #2018-076 to PJ), the FRS-FNRS (Belgium; #T007214 to FL and #J002517 to PJ), the Université catholique de Louvain, the Centre du Cancer (Cliniques universitaires St-Luc), the Fonds Maisin, and the COST Action BM1204 EU_Pancreas. LC, EG and NM hold a Télévie fellowship (#74600115 and #7651017).

*   Competing interests None declared.

*   Provenance and peer review Not commissioned; externally peer reviewed.

*   Author note PJ is senior research associate at the FRS-FNRS (Belgium).

*   Patient consent for publication Not required.

## References

1.   Ying H , Dey P , Yao W , et al . Genetics and biology of pancreatic ductal adenocarcinoma. Genes Dev 2016;30:355–85.[doi:10.1101/gad.275776.115](http://dx.doi.org/10.1101/gad.275776.115) 
    
    [Abstract/FREE Full Text](http://gut.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiZ2VuZXNkZXYiO3M6NToicmVzaWQiO3M6ODoiMzAvNC8zNTUiO3M6NDoiYXRvbSI7czoyMToiL2d1dGpubC82OS80LzcwNC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

2.   Kopp JL , von Figura G , Mayes E , et al . Identification of Sox9-dependent acinar-to-ductal reprogramming as the principal mechanism for initiation of pancreatic ductal adenocarcinoma. Cancer Cell 2012;22:737–50.[doi:10.1016/j.ccr.2012.10.025](http://dx.doi.org/10.1016/j.ccr.2012.10.025) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/j.ccr.2012.10.025&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=23201164&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 
    
    [Web of Science](http://gut.bmj.com/lookup/external-ref?access_num=000312467000007&link_type=ISI) 

3.   Roy N , Hebrok M . Regulation of Cellular Identity in Cancer. Dev Cell 2015;35:674–84.[doi:10.1016/j.devcel.2015.12.001](http://dx.doi.org/10.1016/j.devcel.2015.12.001) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/j.devcel.2015.12.001&link_type=DOI) 

4.   Bardeesy N , Cheng KH , Berger JH , et al . Smad4 is dispensable for normal pancreas development yet critical in progression and tumor biology of pancreas cancer. Genes Dev 2006;20:3130–46.[doi:10.1101/gad.1478706](http://dx.doi.org/10.1101/gad.1478706) 
    
    [Abstract/FREE Full Text](http://gut.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiZ2VuZXNkZXYiO3M6NToicmVzaWQiO3M6MTA6IjIwLzIyLzMxMzAiO3M6NDoiYXRvbSI7czoyMToiL2d1dGpubC82OS80LzcwNC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

5.   Siveke JT , Einwächter H , Sipos B , et al . Concomitant pancreatic activation of Kras(G12D) and Tgfa results in cystic papillary neoplasms reminiscent of human IPMN. Cancer Cell 2007;12:266–79.[doi:10.1016/j.ccr.2007.08.002](http://dx.doi.org/10.1016/j.ccr.2007.08.002) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/j.ccr.2007.08.002&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=17785207&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 
    
    [Web of Science](http://gut.bmj.com/lookup/external-ref?access_num=000249514500009&link_type=ISI) 

6.   Vincent DF , Yan KP , Treilleux I , et al . Inactivation of TIF1gamma cooperates with Kras to induce cystic tumors of the pancreas. PLoS Genet 2009;5:e1000575.[doi:10.1371/journal.pgen.1000575](http://dx.doi.org/10.1371/journal.pgen.1000575) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1371/journal.pgen.1000575&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=19629168&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

7.   von Figura G , Fukuda A , Roy N , et al . The chromatin regulator Brg1 suppresses formation of intraductal papillary mucinous neoplasm and pancreatic ductal adenocarcinoma. Nat Cell Biol 2014;16:255–67.[doi:10.1038/ncb2916](http://dx.doi.org/10.1038/ncb2916) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1038/ncb2916&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=24561622&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

8.   Taki K , Ohmuraya M , Tanji E , et al . GNAS(R201H) and Kras(G12D) cooperate to promote murine pancreatic tumorigenesis recapitulating human intraductal papillary mucinous neoplasm. Oncogene 2016;35:2407–12.[doi:10.1038/onc.2015.294](http://dx.doi.org/10.1038/onc.2015.294) 
    
    

9.   Qiu W , Tang SM , Lee S , et al . Loss of Activin Receptor Type 1B Accelerates Development of Intraductal Papillary Mucinous Neoplasms in Mice With Activated KRAS. Gastroenterology 2016;150:218–28.[doi:10.1053/j.gastro.2015.09.013](http://dx.doi.org/10.1053/j.gastro.2015.09.013) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1053/j.gastro.2015.09.013&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=26408346&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

10.  Yamaguchi J , Mino-Kenudson M , Liss AS , et al . Loss of Trefoil Factor 2 From Pancreatic Duct Glands Promotes Formation of Intraductal Papillary Mucinous Neoplasms in Mice. Gastroenterology 2016;151:1232–44.[doi:10.1053/j.gastro.2016.07.045](http://dx.doi.org/10.1053/j.gastro.2016.07.045) 
    
    

11.  Guerra C , Collado M , Navas C , et al . Pancreatitis-induced inflammation contributes to pancreatic cancer by inhibiting oncogene-induced senescence. Cancer Cell 2011;19:728–39.[doi:10.1016/j.ccr.2011.05.011](http://dx.doi.org/10.1016/j.ccr.2011.05.011) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/j.ccr.2011.05.011&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=21665147&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 
    
    [Web of Science](http://gut.bmj.com/lookup/external-ref?access_num=000291839500007&link_type=ISI) 

12.  Amato E , Molin MD , Mafficini A , et al . Targeted next-generation sequencing of cancer genes dissects the molecular profiles of intraductal papillary neoplasms of the pancreas. J Pathol 2014;233:217–27.[doi:10.1002/path.4344](http://dx.doi.org/10.1002/path.4344) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1002/path.4344&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=24604757&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

13.  Tan MC , Basturk O , Brannon AR , et al . GNAS and KRAS Mutations Define Separate Progression Pathways in Intraductal Papillary Mucinous Neoplasm-Associated Carcinoma. J Am Coll Surg 2015;220:845–54.[doi:10.1016/j.jamcollsurg.2014.11.029](http://dx.doi.org/10.1016/j.jamcollsurg.2014.11.029) 
    
    

14.  Lee JH , Kim Y , Choi JW , et al . KRAS, GNAS, and RNF43 mutations in intraductal papillary mucinous neoplasm of the pancreas: a meta-analysis. Springerplus 2016;5:1172.[doi:10.1186/s40064-016-2847-4](http://dx.doi.org/10.1186/s40064-016-2847-4) 
    
    

15.  Sato N , Rosty C , Jansen M , et al . STK11/LKB1 Peutz-Jeghers gene inactivation in intraductal papillary-mucinous neoplasms of the pancreas. Am J Pathol 2001;159:2017–22.[doi:10.1016/S0002-9440(10)63053-2](http://dx.doi.org/10.1016/S0002-9440(10)63053-2) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/S0002-9440(10)63053-2&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=11733352&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 
    
    [Web of Science](http://gut.bmj.com/lookup/external-ref?access_num=000172457400007&link_type=ISI) 

16.  Giardiello FM , Brensinger JD , Tersmette AC , et al . Very high risk of cancer in familial Peutz-Jeghers syndrome. Gastroenterology 2000;119:1447–53.[doi:10.1053/gast.2000.20228](http://dx.doi.org/10.1053/gast.2000.20228) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1053/gast.2000.20228&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=11113065&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 
    
    [Web of Science](http://gut.bmj.com/lookup/external-ref?access_num=000165833800007&link_type=ISI) 

17.  Resta N , Pierannunzio D , Lenato GM , et al . Cancer risk associated with STK11/LKB1 germline mutations in Peutz-Jeghers syndrome patients: results of an Italian multicenter study. Dig Liver Dis 2013;45:606–11.[doi:10.1016/j.dld.2012.12.018](http://dx.doi.org/10.1016/j.dld.2012.12.018) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/j.dld.2012.12.018&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=23415580&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

18.  Augereau C , Collet L , Vargiu P , et al . Chronic pancreatitis and lipomatosis are associated with defective function of ciliary genes in pancreatic ductal cells. Hum Mol Genet 2016;25:5017–26.[doi:10.1093/hmg/ddw332](http://dx.doi.org/10.1093/hmg/ddw332) 
    
    

19.  Prévot PP , Augereau C , Simion A , et al . Let-7b and miR-495 stimulate differentiation and prevent metaplasia of pancreatic acinar cells by repressing HNF6. Gastroenterology 2013;145:668–78.[doi:10.1053/j.gastro.2013.05.016](http://dx.doi.org/10.1053/j.gastro.2013.05.016) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1053/j.gastro.2013.05.016&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=23684747&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

20.  Strobel O , Rosow DE , Rakhlin EY , et al . Pancreatic duct glands are distinct ductal compartments that react to chronic injury and mediate Shh-induced metaplasia. Gastroenterology 2010;138:1166–77.[doi:10.1053/j.gastro.2009.12.005](http://dx.doi.org/10.1053/j.gastro.2009.12.005) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1053/j.gastro.2009.12.005&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=20026066&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 
    
    [Web of Science](http://gut.bmj.com/lookup/external-ref?access_num=000275109900046&link_type=ISI) 

21.  Yamaguchi J , Liss AS , Sontheimer A , et al . Pancreatic duct glands (PDGs) are a progenitor compartment responsible for pancreatic ductal epithelial repair. Stem Cell Res 2015;15:190–202.[doi:10.1016/j.scr.2015.05.006](http://dx.doi.org/10.1016/j.scr.2015.05.006) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/j.scr.2015.05.006&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=26100232&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

22.  Shields DJ , Murphy EA , Desgrosellier JS , et al . Oncogenic Ras/Src cooperativity in pancreatic neoplasia. Oncogene 2011;30:2123–34.[doi:10.1038/onc.2010.589](http://dx.doi.org/10.1038/onc.2010.589) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1038/onc.2010.589&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=21242978&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 
    
    [Web of Science](http://gut.bmj.com/lookup/external-ref?access_num=000290249600004&link_type=ISI) 

23.  Delgiorno KE , Hall JC , Takeuchi KK , et al . Identification and manipulation of biliary metaplasia in pancreatic tumors. Gastroenterology 2014;146:233–44.[doi:10.1053/j.gastro.2013.08.053](http://dx.doi.org/10.1053/j.gastro.2013.08.053) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1053/j.gastro.2013.08.053&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=23999170&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 
    
    [Web of Science](http://gut.bmj.com/lookup/external-ref?access_num=000328736000037&link_type=ISI) 

24.  Wei D , Wang L , Yan Y , et al . KLF4 Is Essential for Induction of Cellular Identity Change and Acinar-to-Ductal Reprogramming during Early Pancreatic Carcinogenesis. Cancer Cell 2016;29:324–38.[doi:10.1016/j.ccell.2016.02.005](http://dx.doi.org/10.1016/j.ccell.2016.02.005) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/j.ccell.2016.02.005&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=26977883&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

25.  David CJ , Huang YH , Chen M , et al . TGF-β Tumor Suppression through a Lethal EMT. Cell 2016;164:1015–30.[doi:10.1016/j.cell.2016.01.009](http://dx.doi.org/10.1016/j.cell.2016.01.009) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/j.cell.2016.01.009&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=26898331&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

26.  Daley D , Zambirinis CP , Seifert L , et al . gammadelta T Cells Support Pancreatic Oncogenesis by Restraining alphabeta T Cell Activation. Cell 2016;166:1485–99.
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/j.cell.2016.07.046&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=27569912&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

27.  Ferreira RMM , Sancho R , Messal HA , et al . Duct- and Acinar-Derived Pancreatic Ductal Adenocarcinomas Show Distinct Tumor Progression and Marker Expression. Cell Rep 2017;21:966–78.[doi:10.1016/j.celrep.2017.09.093](http://dx.doi.org/10.1016/j.celrep.2017.09.093) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1016/j.celrep.2017.09.093&link_type=DOI) 

28.  Kopp JL , Dubois CL , Schaeffer DF , et al . Loss of Pten and Activation of Kras Synergistically Induce Formation of Intraductal Papillary Mucinous Neoplasia From Pancreatic Ductal Cells in Mice. Gastroenterology 2018;154.[doi:10.1053/j.gastro.2017.12.007](http://dx.doi.org/10.1053/j.gastro.2017.12.007) 
    
    

29.  Wu J , Jiao Y , Dal Molin M , et al . Whole-exome sequencing of neoplastic cysts of the pancreas reveals recurrent mutations in components of ubiquitin-dependent pathways. Proc Natl Acad Sci U S A 2011;108:21188–93.[doi:10.1073/pnas.1118046108](http://dx.doi.org/10.1073/pnas.1118046108) 
    
    [Abstract/FREE Full Text](http://gut.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NDoicG5hcyI7czo1OiJyZXNpZCI7czoxMjoiMTA4LzUyLzIxMTg4IjtzOjQ6ImF0b20iO3M6MjE6Ii9ndXRqbmwvNjkvNC83MDQuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

30.  Furukawa T , Kuboki Y , Tanji E , et al . Whole-exome sequencing uncovers frequent GNAS mutations in intraductal papillary mucinous neoplasms of the pancreas. Sci Rep 2011;1:161.[doi:10.1038/srep00161](http://dx.doi.org/10.1038/srep00161) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1038/srep00161&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=22355676&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 

31.  Castellone MD , Teramoto H , Williams BO , et al . Prostaglandin E2 promotes colon cancer cell growth through a Gs-axin-beta-catenin signaling axis. Science 2005;310:1504–10.[doi:10.1126/science.1116221](http://dx.doi.org/10.1126/science.1116221) 
    
    [Abstract/FREE Full Text](http://gut.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6Mzoic2NpIjtzOjU6InJlc2lkIjtzOjEzOiIzMTAvNTc1My8xNTA0IjtzOjQ6ImF0b20iO3M6MjE6Ii9ndXRqbmwvNjkvNC83MDQuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

32.  Koo BK , Spit M , Jordens I , et al . Tumour suppressor RNF43 is a stem-cell E3 ligase that induces endocytosis of Wnt receptors. Nature 2012;488:665–9.[doi:10.1038/nature11308](http://dx.doi.org/10.1038/nature11308) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1038/nature11308&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=22895187&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 
    
    [Web of Science](http://gut.bmj.com/lookup/external-ref?access_num=000308095100058&link_type=ISI) 

33.  Ghaleb AM , Nandan MO , Chanchevalap S , et al . Krüppel-like factors 4 and 5: the yin and yang regulators of cellular proliferation. Cell Res 2005;15:92–6.[doi:10.1038/sj.cr.7290271](http://dx.doi.org/10.1038/sj.cr.7290271) 
    
    [CrossRef](http://gut.bmj.com/lookup/external-ref?access_num=10.1038/sj.cr.7290271&link_type=DOI) 
    
    [PubMed](http://gut.bmj.com/lookup/external-ref?access_num=15740636&link_type=MED&atom=%2Fgutjnl%2F69%2F4%2F704.atom) 
    
    [Web of Science](http://gut.bmj.com/lookup/external-ref?access_num=000227709800002&link_type=ISI) 

34.  Katz JP , Perreault N , Goldstein BG , et al . The zinc-finger transcription factor Klf4 is required for terminal differentiation of goblet cells in the colon. Development 2002;129:2619–28.
    
    [Abstract/FREE Full Text](http://gut.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoiZGV2ZWxvcCI7czo1OiJyZXNpZCI7czoxMToiMTI5LzExLzI2MTkiO3M6NDoiYXRvbSI7czoyMToiL2d1dGpubC82OS80LzcwNC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

35.  Song E , Ma X , Li H , et al . Attenuation of krüppel-like factor 4 facilitates carcinogenesis by inducing g1/s phase arrest in clear cell renal cell carcinoma. PLoS One 2013;8:e67758.[doi:10.1371/journal.pone.0067758](http://dx.doi.org/10.1371/journal.pone.0067758)