A novel mouse model of intrahepatic cholangiocarcinoma induced by liver-specific Kras activation and Pten deletion

Intrahepatic cholangiocarcinoma (ICC) is an aggressive malignancy with poor prognosis and its incidence is increasing worldwide. Recently, several types of cells have been considered as the origin of ICC, namely cholangiocytes, liver progenitor cells, and hepatocytes. Here, we have established a novel mouse model of ICC by liver-specific Kras activation and Pten deletion. An activating mutation of Kras in combination with deletion of Pten was introduced in embryonic hepatic bipotential progenitor cells (so-called hepatoblasts) and mature hepatocytes using the Cre-loxP system. As a result, liver-specific Kras activation and homozygous Pten deletion cooperated to induce ICCs exclusively. In contrast, Kras activation in combination with heterozygous Pten deletion induced both ICCs and HCCs, whereas Kras activation alone resulted in HCCs but not ICCs. Furthermore, a cell-lineage visualization system using tamoxifen-inducible Cre-loxP demonstrated that the ICCs did not originate from hepatocytes but from cholangiocytes. Our data suggest that mice carrying liver-specific Kras activation in combination with homozygous Pten deletion should be useful for the investigation of therapeutic strategies for human ICC.

on chromosome 10 (PTEN), a phosphoprotein/phospholipid dual-specificity phosphatase that preferentially dephosphorylates the second-messenger molecule phosphatidylinositol-3,4,5-trisphosphate, to antagonize the activity of PI3K. Loss of PTEN function also plays a vital role in a wide range of tumors through the constitutive activation of the PI3K-AKT pathway. Importantly, PTEN mutations were observed in ~11% of ICCs 8,17 , and inhibitory phosphorylation of PTEN has been reported in ICCs 18 . Hypermethylation of the CpG island in the PTEN promoter region was also reported in liver fluke-related ICCs 19 . In addition, microRNA-21 that inhibits PTEN is frequently overexpressed in ICCs 20 . These data suggest that inactivation of PTEN is one of the causes of activated PI3K/AKT signaling in human ICC.
Genetically engineered mouse models (GEMs) are essential tools to study the molecular pathogenesis of human diseases and assess novel therapeutic strategies preclinically. As a model of human pancreatic cancer, mice with pancreas-specific Kras activation were generated. The mice slowly developed premalignant pancreatic intraepithelial neoplasms (PanIN) with prolonged latencies and incomplete penetrance for the development of pancreatic ductal adenocarcinoma (PDAC) 21 . Furthermore, other models of rapidly progressing PDAC have been established by crossing the mice carrying pancreas-specific Kras activation with mice carrying loss of function in other genes that are associated with PDAC, such as Ink4a/Arf, Trp53, Tgf β type II receptor, or Smad4 [22][23][24][25] .
In contrast to PDAC, a limited number of GEMs for ICC have been reported. These models include liver-specific Pten and Smad4 double knockout, Kras activation in combination with Trp53 knockout, inactivation of Hippo pathway by knockout of Sav1, and induction of intracellular domain of Notch receptor (NICD) 18,[26][27][28] . It is noteworthy that most of these mouse models developed HCC as well as ICC because the genetic engineering was introduced in both hepatocytes and bipotential progenitor cells.
In the present study, we have established a novel genetically engineered mouse model of ICC. We found that Kras activation and homozygous Pten deletion in embryonic bipotential progenitor cells cooperate to exclusively induce ICC reminiscent of human ICC. Although several recent reports suggest the possibility that ICC can arise from hepatocytes through Notch-mediated transdifferentiation 29,30 , we have demonstrated, using cholangiocyteor hepatocyte-specific Cre-loxP system, that ICCs in this model originated from cholangiocytes that were differentiated from Kras activated and Pten deleted progenitor cells.

Kras activation and Pten deletion in hepatoblasts induce intrahepatic cholangiocarcinoma.
To investigate the role of cooperation of Kras activation and Pten inactivation in hepatotumorigenesis in vivo, we took advantage of conditional activation of a mutant Kras (LSL-Kras G12D ) allele to express oncogenic Kras at a physiological level under endogenous Kras promoter after Cre-mediated recombination 31 , and conditional knockout allele of Pten (Pten flox ) 32 . We crossed mice carrying a LSL-Kras G12D allele and/or a Pten flox allele with mice expressing Cre recombinase under the control of albumin promoter (Alb-Cre) 33 to express the oncogenic Kras G12D allele and/or to delete Pten alleles specifically in the liver (Fig. 1A).
We first analyzed macroscopic phenotype of liver and survival in Alb-Cre + ; LSL-Kras G12D/+ ; Pten flox/flox mice (AKPP mice), Alb-Cre + ; LSL-Kras G12D/+ ; Pten flox/+ mice (AKP mice), and Alb-Cre + ; LSL-Kras G12D/+ ; Pten +/+ mice (AK mice). In these mice, both Pten alleles (AKPP mice), one Pten allele (AKP mice), and no Pten allele (AK mice) was deleted in combination with Kras activation specifically in the liver. Mice with these genotypes were obtained at the expected Mendelian frequency at birth but the AKPP mice alone showed apparent growth retardation. All of the AKPP mice started to demonstrate abdominal distension at ~5 weeks of age, frequently accompanied by jaundice and weight loss. The distension was caused by hepatic enlargement and/or hemorrhagic ascites. These manifestations recapitulated well those frequently observed in human ICC. They were euthanized when they appeared to be in distress. Autopsies revealed the presence of multiple solid tumors with various sizes throughout the liver (Fig. 1B-D). The median survival was 46 days (Fig. 1E).
AKP and AK mice also developed liver tumors at 6 and 12 months, respectively (Fig. 1E). Although the majority of AKP mice did not show obvious symptoms until 7 months of age, they suffered from abdominal distension due to multiple liver tumors at later stages. On the other hand, AK mice did not show obvious symptoms up to an age of 18 months (Fig. 1E).
We next evaluated histological changes of the liver tumors in AKPP mice. Although no apparent abnormality was histologically found in the liver of AKPP mice until 3 weeks of age (Fig. 1F), various degrees of hyperplasia in the bile ducts, characterized by increase in number and size, and change in morphology, were observed at 4 weeks of age. At 5 weeks, a part of the hyperplastic ductal lesions at the hilum became enlarged and showed the pattern of papillary growth, and the number of dysplastic ducts was increased (Fig. 1G). After 7 weeks, invasive tumors were apparent with an abundant desmoplastic stroma, or "desmoplasia" (Fig. 1H). The tumors primarily showed glandular morphology that resembled well-differentiated human CC. In addition, regions of moderately differentiated tumor with a cribriform appearance were observed in some cases (Fig. 1H,I). No apparent metastasis or invasion to other organs was observed in AKPP mice.
The presence of desmoplasia around the tumor is generally considered to be one of the most characteristic histological findings for distinguishing ICC from HCC 34 . The Masson-Trichrome staining frequently depicted tumors surrounded by dense fibrous stroma in AKPP mice, indicating the accumulation of fibrillar collagens ( Fig. 2A). Immunohistochemical analysis with an anti-α SMA antibody showed positive staining in the stroma surrounding the tumors, which implied that the stellate cells were activated and acquired a myofibroblast-like phenotype (Fig. 2B). Alcian blue staining exhibited mucin production in the tumor cells (Fig. 2C), suggesting that they have characteristics of epithelial cells in the bile duct.
To further clarify the characteristics of the tumor cells developed in AKPP mice, we analyzed various markers used for diagnosis of ICC. The tumor cells showed positive staining for pan-cytokeratin (pan-Ck) and Ck19 (Fig. 2D,E) 35 , markers of cholangiocyte differentiation. On the other hand, they were negative for Hnf4α , a specific marker of hepatocyte differentiation (Fig. 2F) 35 . These results demonstrated that Kras activation and Pten homozygous deletion cooperate to induce ICC exclusively.
To confirm the activation of Mapk and Pi3k pathways in the tumors of the AKPP mice, immunohistochemical staining of p-Erk and p-Akt was conducted. As expected, the tumor cells were positive for both p-Erk and p-Akt ( Fig. 2G,H). In addition, immunohistochemical staining of p-S6, a marker for the activation of mTorc1, was positive in the tumor cells (Fig. 2I). These results suggested that Mapk and Pi3k-mTorc1 pathways were activated in the ICC induced by Kras activation and Pten deletion.
In contrast to the tumors in AKPP mice, AKP mice developed liver tumors with both hepatocyte and cholangiocyte differentiation. Histological examination revealed that the majority of large nodules in the AKP mice were hepatocellular dysplasia (HD), which is considered as a precursor lesion of HCC ( Figure S1A). Partially, there were lesions diagnosed as ICC by the appearance of adenocarcinoma showing tubular and/or papillary structures with a fibrous stroma ( Figure S1B). Liver parenchyma surrounding these tumors was not cirrhotic. Interestingly, immunohistochemical staining of p-Erk and p-Akt revealed the activation of Mapk pathway, but not the activation of Pi3k pathway, in HDs ( Figure S2A,C), whereas coactivation of these two pathways was detected in ICC lesions ( Figure S2B,D). In addition, loss of Pten expression was shown in the tumor cells of ICCs in AKP mice by immunohistochemical staining (Figure S3A,B), suggesting that the remaining wild-type Pten allele was inactivated by some mechanism such as loss of heterozygosity, loss-of-function mutation, and hypermethylation of its promoter region.
Regarding AK mice, they showed no abnormality until 12 months of age, but they later exhibited abdominal distension due to the enlarged liver with multiple tumors. Histopathological examination revealed that the tumors resembled HD that developed in the AKP mice ( Figure S1C). No abnormality in biliary system was detected in the liver of the AK mice.
Collectively, in combination with oncogenic Kras expression, Pten dose might dictate the fate determination of hepatotumorigenesis into biliary or hepatocyte cell lineage. In addition, it is noteworthy that homozygous Pten deletion in cooperation with oncogenic Kras mutation induces ICC exclusively.
Kras activation and Pten deletion in cholangiocytes but not hepatocytes induce ICC. To elucidate whether mature hepatocytes can be the origin of ICC induced by Kras activation and Pten deletion, we crossed LSL-Kras G12D/+ ; Pten flox/flox mice with Alb-CreERT2 + mice in which tamoxifen (TMX)-inducible CreERT2 was knocked into endogenous Alb locus 36 . These Alb-CreERT2 + ; LSL-Kras G12D/+ ; Pten flox/flox mice (A ER KPP mice) were injected with TMX at eight weeks after birth and tumor development was investigated three months later ( Fig. 3A). Consequently, multiple liver tumors were observed in these mice (Fig. 3B). Histological analyses showed that all tumors were HCC and HD but not ICC (Fig. 3C). These results suggest that Kras activation and Pten deletion in differentiated hepatocytes induce HCC but not ICC.
We next administered TMX to A ER KPP mice at postnatal day 10 (P10) to investigate the possibility that hepatocytes in young mice could be the origin of ICC. Interestingly, these mice developed ICCs exclusively, which phenocopied the AKPP mice (Fig. 3D,E).
The fact that the different timing of TMX administration altered the types of liver tumor prompted us to examine the possibility of the age-dependent difference of cell types where Cre-mediated recombination occurred in the A ER KPP mice. To test this possibility, we generated Alb-CreERT2 + ; Rosa26-mTmG reporter (Alb-CreERT2 + ; R26R mTmG/+ ) mice by crossing Alb-CreERT2 mice with R26R mTmG mice that express membrane-targeted EGFP (mG) in Cre-expressing cells and membrane-targeted tdTomato (mT) in Cre-absent cells 37 . Examination of the liver in Alb-CreERT2 + ; R26R mTmG/+ mice with TMX-administration at eight weeks after birth depicted EGFP expression only in the hepatocytes (Fig. 3F). In contrast, in the mice administered with TMX at P10, EGFP expression was detected not only in the hepatocytes but also in the cells that form the bile ducts (Fig. 3G). These results suggest that ICCs originated from the cholangiocytes carrying Cre-mediated Kras activation and Pten deletion in the A ER KPP mice.

Kras activation and Pten deletion in mature choalngiocytes induce ICC. To investigate whether
Kras activation and Pten deletion in intrahepatic bile ducts induce ICC, we took advantage of mice with a TMX-inducible Cre under the control of Ck19 promoter (K19 CreERT ) 38 . We crossed them with LSL-Kras G12D mice and Pten flox mice to generate K19 CreERT/+ ; LSL-Kras G12D/+ ; Pten flox/flox mice (K ER KPP mice). We administered TMX at eight weeks of age (Fig. 4A). The median survival of the K ER KPP mice was 30 days after TMX injection. Although no obvious tumors were macroscopically observed in the liver, their extrahepatic bile duct as well as the gallbladder and cystic duct were rigid and dilated (Fig. 4B). Histological analysis of the liver revealed pre-malignant papillary ductal lesions in periportal areas (Fig. 4C). Besides the ductal lesions in their liver, a variety of organs showed pathological abnormalities, including papillary hyperplasia of the extrahepatic biliary tract and gallbladder, PanIN-like lesions and hyperplasia in the pancreatic ducts, hyperplastic changes of gastric mucosa, colonic serrated epithelial changes, and papillary hyperplasia of bronchial epithelia that accompanied obstructive pneumonia (Fig. 4D,E and Figure S4A,B). Respiratory failure by the lung lesions was most likely to be the cause of death of the K ER KPP mice. The hepatobiliary phenotypes in the K ER KPP mice were consistent with those reported in Ah-CreER T/+ ; LSL-Kras G12V/+ ; Pten flox/flox mice, in which β -naphthoflavone/TMX-inducible Kras activation and Pten deletion were introduced in biliary as well as gastrointestinal epithelia 39 . The premalignant ductal lesions in the liver were positive for pan-Ck (Fig. 4F), confirming that Kras activation and Pten deletion is capable of inducing ICCs from cholangiocytes. In addition, we observed positive immunohistochemical staining with both P-Erk and P-Akt antibodies in the neoplastic cells (Fig. 4G,H), suggesting that Mapk and Pi3k pathways were activated by the activation of Kras and the deletion of Pten.

Discussion
In this study, we have established a novel mouse model of ICC by liver-specific Kras activation and Pten deletion (AKPP). The AKPP mice, in which Kras activation and homozygous deletion of Pten were introduced specifically in the liver, developed exclusively ICCs. There have been several reports of genetically engineered mouse models of ICC but most of the models develop HCC as well as ICC 18,26,27 . Similarly, the AKP mice, in which Pten allele was heterozygously inactivated, developed both HCC and ICC. It has been reported that liver-specific homozygous deletion of Pten induces HCC as well as ICC in mice at more than 12 months of age 18,40 . These reports suggest two possible explanations for our observation that the AKPP mice exclusively developed ICCs. One is that the limited life-span of the AKPP mice (median survival: 46 days) was not long enough to develop HCCs. The other is that excessive activation of Pi3k pathway in combination with oncogenic Kras expression promotes the formation of ICC, but not the formation of HCC. In agreement with the latter possibility, immunohistochemical analysis revealed that the activation of Pi3k pathway was observed in the ICCs but not in HCCs/HDs in the AKP mice. Since Pten expression was lost in the tumor cells of ICCs in the AKP mice, ICCs were induced in the mice through the biallelic inactivation of Pten in combination with Kras activation as observed in the AKPP mice. It is of note that the AK mice with oncogenic Kras along with intact Pten expression in the liver developed HCCs/ HDs but not ICCs. Therefore, the activation of Pi3k pathway might determine the fate of tumorous cells caused by oncogenic Kras from hepatocyte linage to biliary lineage.
Investigating the cell origin of ICC has been considered to be important for the development of new therapies for this devastating disease. ICC has been recently classified into three groups, pure mucin-producing ICC (muc-ICC), predominant mucin-producing ICC with hepatocytic differentiation areas and/or ductular areas in the tumor (mixed-ICC), and predominant ductular areas with hepatocytic differentiation areas with muc-producing ICC areas (cholangiolocellular carcinoma; CLC) 41 . Muc-ICCs are thought to be of cholangiocyte origin, whereas mixed-ICC and CLCs are thought to originate from hepatic progenitor cells 41 , by their clinicopathological and molecular features. Meanwhile, several groups have reported that ICCs can originate from differentiated hepatocytes by transdifferentiation 29,30 . In the present study, we have shown that the ICCs were induced by liver-specific Kras G12D expression and Pten deletion using Alb-Cre mice, Alb-CreERT2 mice with TMX-administration at P10, and K19 CreERT mice with TMX-administration at P42. In contrast, no ICCs were induced using Alb-CreERT2 mice with TMX-administration at P42. TMX-administration at P10 was shown to cause Cre-loxP-mediated gene recombination in bile duct cells as well as hepatocytes, whereas TMX-administration at P42 caused the recombination only in hepatocytes. Taken together, these results suggest that cholangiocytes are likely to be the origin of ICC in these models. In agreement with our results, Guest et al. have recently revealed that biliary epithelia is a possible origin of ICC using cholangiocyte-lineage tracing system to induce p53 loss in biliary epithelia in combination with chronic liver injury 42 . Although the ICCs in the AKPP mice were originated from cholangiocytes like muc-ICCs in human, they do not so frequently produce mucin as muc-ICCs. This finding suggests that Kras activation and Pten deletion are not enough to the induction of mucin production in biliary epithelia. Alternatively, murine biliary epithelia might be insensitive to intracellular signaling that promote mucin production. Nevertheless, the AKPP mice should be a useful model for the investigation of molecular mechanisms of the development of ICC from cholangiocytes.
Comprehensive genome profiling of intrahepatic cholangiocarcinoma has been recently reported by several groups 8,12,[43][44][45][46] . These studies revealed that TP53 and KRAS are two of the genes most frequently mutated in human ICC. In addition, Zou et al. revealed that the RAS/PI3K pathway-related genes were frequently altered by the pathway analysis of 103 ICC patients 8 . Consistent with these data, an inactivating mutation of PTEN or an activating mutation of PIK3CA, the two representative mutations that activate the PI3K pathway, was shown to coexist with oncogenic KRAS mutation in a subset of human ICCs 8 . The AKPP mice should be the most suitable mouse model to facilitate the development of new therapies for ICCs with an oncogenic KRAS mutation and the activated PI3K pathway.
In conclusion, we have established a novel mouse model of ICC by introducing the oncogenic Kras mutation and the homozygous Pten deletion in the liver. The tumors induced in this model were exclusively ICCs that histologically recapitulate human ICC of cholangiocyte origin. Intensive investigation of this model should provide better understanding of human ICCs derived from biliary epithelial cells and facilitate the development of new therapies for this devastating disease.

Materials and Methods
Mice. Alb-Cre, LSL-Kras G12D , and R26R mTmG mice were purchased from The Jackson Laboratory 31,33,37 . Pten flox , Alb-CreERT2, and K19 CreERT mice were kindly provided by Tak W Mak (University of Toronto, Toronto, Canada), Daniel Metzger and Pierre Chambon (IGBMC, Illkirch-Cedex, France), and Guoqiang Gu (Vanderbilt University, Nashville, TN), respectively 32,36,38 . All mice were on the C57BL/6 genetic background. PCR primers used for the genotyping of each mouse strain are listed in Supplemental Table S1. All mice were housed in specific pathogen-free conditions within the animal care facility at the University of Tokyo. All the experimental protocols were approved by the Ethics Committee for Animal Experimentation and conducted in accordance with the Guidelines for the Care and Use of Laboratory Animals of the Institute of Medical Science and Department of Medicine, the University of Tokyo.
Tamoxifen treatment of mice. For the induction of tumors in the A ER KPP mice, the mice were treated with 1 mg and 5 mg of tamoxifen (TMX) at P10 and P56, and then euthanized 2months and 3 months later, respectively. For tumor induction in the K ER KPP mice, the mice were treated with 5 mg of TMX at P56 and then euthanized one month later.
To analyze the origin of tumors in the A ER KPP mice, the Alb-CreERT2 mice were treated with 1 mg and 5 mg of TMX at P10 and P56 and then euthanized one week later.
Histology and immunohistochemistry. Normal and tumor tissues were fixed in 4% paraformaldehyde overnight then processed, paraffin-embedded, sectioned and stained with hematoxylin and eosin, Masson-trichrome, and Alcian blue according to standard protocols. For immunohistochemistry, 5 μm sections were incubated with primary antibodies overnight at 4 °C in a humidified chamber. The following primary antibodies were used: HNF4α (Santa Cruz), CK19 (TROMAIII, Developmental Hybrydoma Bank), Pan-cytokeratin (AE1/AE3, GeneTex), p-AKT Ser473 (Cell Signaling), p-ERK (Cell Signaling), p-S6 (Cell Signaling), PTEN (Cell Signaling), and α SMA (Sigma). For rabbit and rat antibodies, the sections were subsequently developed using a VECTASTAIN Elite ABC kit (Vector Laboratories). For mouse monoclonal staining, a MOM kit (Vector Laboratories) was also used. Representative sections from at least three mice were counted for each genotype. For immunofluorescence, liver tissues with mTmG expression were fresh-frozen in Optimal Cutting Temperature medium (Tissue-Tek). 5 μm sections were fixed with methanol for 10 min at −20 °C, washed 3 times with PBS, mounted in VECTASHIELD mounting medium with DAPI (Vector Laboratories), and imaged with confocal microscopy (Nikon).