Zebrafish model of KRAS-initiated pancreatic cancer

ABSTRACT Pancreatic cancer constitutes a genetic disease in which somatic mutations in the KRAS proto-oncogene are detected in a majority of tumors. KRAS mutations represent an early event during pancreatic tumorigenesis that crucial for cancer initiation and progression. Here, we established a zebrafish pancreatic cancer model that highly recapitulates human pancreatic intraepithelial neoplasia (PanIN) development. We established a novel system combining CRE/Lox technology with the GAL4/UAS system to express oncogenic KRAS in the ptf1a domain temporarily. In this system, zebrafish developed PanIN at an 11.1% rate by 24 and 36 weeks after KRASG12V induction. The histological and immunohistochemical profiles of these experimental tumors bore striking resemblance to human PanIN. Within the whole abnormal area, the entire spectrum of differentiation ranging from PanIN-1 to PanIN-3 was noted. Immunohistochemical analysis including Alcian blue, CK-18, cadhedrin-1, and DCLK1 staining confirmed the PanIN region as a characteristic pancreatic cancer precursor lesion. Taken together, these findings demonstrate that this zebrafish model may offer the possibility of an experimental and preclinical system to evaluate different strategies for targeting pancreatic tumors and finally improve the outcome for the patients with pancreatic tumors.


Introduction
Pancreatic cancer is the fourth leading cause of cancer death in the United States, with a very low rate of survival following diagnosis (Jemal et al. 2008). Most pancreatic cancers comprise infiltrating pancreatic ductal adenocarcinomas (PDAC), which is considered to arise from ductal precursor lesions termed pancreatic intraepithelial neoplasia (PanIN) . In humans, 45% of early PanIN lesions and 90% of infiltrating PDAC harbor oncogenic KRAS mutations (Hruban et al. 1993;Hruban et al. 2000;Jones et al. 2008). Several mouse models for KRAS-mediated pancreatic cancer have been generated. A major breakthrough emerged from an approach targeting an oncogenic KRAS in presumed progenitor cells, using pdx1 or ptf1a drivers (Hingorani et al. 2003). These mutant mice developed PanIN lesions similar to those seen in humans, underscoring the fact that oncogenic KRAS functions as a critical initiator of pancreatic tumorigenesis (Hingorani et al. 2003). The dramatic progress of mouse models for pancreatic cancer arouses the expectation that these models could be used to develop a new drug against PDAC, as preclinical trials of anticancer drugs in animal models constitute an important step in the drug development (Aguirre et al. 2003;Hingorani et al. 2003;Hruban et al. 2006). However, the findings of most preclinical trials of the new drugs that showed a good efficacy against PDAC in a mouse model were not confirmed during the clinical phase (Kapischke and Pries 2008). Addressing this fundamental problem will likely require the development of a new platform that could be highly utilized in drug development.
The zebrafish has emerged as an excellent model organism for the study of human cancer. Zebrafish show the same characteristics as human cancers, such as genomic instability, invasiveness, and metastasis (Le et al. 2007). Moreover, the optical transparency of the zebrafish embryo makes it possible to track the locations or activities of genes of interest (Beis and Stainier 2006;Liu et al. 2008). For example, tagging the transgene with a fluorescent protein allows for the real-time observation of tumor progression and invasion. It also makes it possible to monitor drug efficacy in vivo and to assess the severity of a specific phenotype visually. In our previous studies, we established two KRAS-initiated pancreatic cancer models in zebrafish (Park et al. 2008;Liu and Leach 2011). Our first model used a genomic bacterial artificial chromosome (BAC) (CH211-142H2) spanning the zebrafish ptf1a locus to express the human KRAS G12V mutant in the exocrine pancreas (Park et al. 2008). After oncogenic KRAS was expressed in developing zebrafish pancreas, pancreatic progenitor cells failed to undergo normal exocrine differentiation, leading to the subsequent formation of invasive pancreatic cancer. However, this model induced predominant acinar cell carcinomas as opposed to the classical pancreatic cancer development (Park et al. 2008). Our second model used the GAL4/UAS system to express the human KRAS G12V mutant in the exocrine pancreas (Liu and Leach 2011). Instead of expressing the human KRAS G12V mutant directly from ptf1a regulatory elements, we used a novel GAL4/UAS approach that allows the simultaneous expression of human KRAS G12V mutant in ptf1a-expressing pancreatic cell types. However, this model also induced predominant acinar cell carcinomas with a few tumors with mixed acinar and ductal differentiation (Liu and Leach 2011).
In the present study, we combined CRE/Lox and GAL4/UAS systems to establish the first KRAS-initiated pancreatic cancer model in zebrafish that highly recapitulates human PanIN development. As this model showed the whole spectrum of PanINs, it holds promise as an experimental model system to evaluate different strategies for targeting pancreatic cancers.

Results
Identification of the PanIN regions in Tg (ptf1a: CRE ERT2 ; LSL-GAL4; UAS-KRAS G12V ) fish To control KRAS G12V expression in a tissue-specific manner, we generated a conditional GAL4-VP16 transactivator under the control of the ubiquitin b (ubb) promoter with a lox-stop-lox (LSL) cassette inserted between the promoter and the GAL4-VP16, Tg (ubb:LSL:GAL4-VP16) (herein LSL-GAL4). Next, we crossed the Tg (LSL-GAL4) line with the Tg (UAS:eGFP-KRAS G12V ) (herein UAS-KRAS G12V ) line to direct eGFP-KRAS G12V expression (Liu et al. 2011). To activate the KRAS gene in zebrafish pancreatic progenitor cells, we used the inducible Tg (ptf1a:CRE ERT2 ) line, which induces the expression of the transgene in ptf1a-expressing cells of the zebrafish pancreas after 4-OHT treatment (Wang et al. 2015). Finally, we established a triple transgenic fish Tg (ptf1a:CRE ERT2 : LSL-GAL4: UAS-KRAS G12V ). This novel activation system induced the expression of GAL4 under the control of the ubb promoter after excision of an LSL cassette by CRE in the ptf1a domain (Figure 1(A)). In turn, GAL4 proteins bind to their specific recognition sequence, upstream activating sequence (UAS), and stimulate transcription of oncogenic KRAS G12V (Figure 1(A)).
To induce the expression of eGFP-KRAS G12V , triple transgenic fish Tg (ptf1a:CRE ERT2 ; LSL-GAL4: UAS-KRAS G12V ) were treated with 5 µM 4-OHT in the E3 medium. A random subset of fish was anesthetized periodically and sacrificed at 12-, 24-, and 36-week time points after 4-OHT treatment. At 12 weeks after 4-OHT treatment, all examined fish showed histologically normal pancreas and no evidence of tumor formation in any organ (n = 9, data not shown). At 24 and 36 weeks after 4-OHT treatment, 1/9 fish (11.1%) developed PanIN, which recapitulated human PanIN (Figure 1(B)). eGFP fluorescence in the pancreas was observed from the dissected abdominal viscera of zebrafish that developed PanIN. This fluorescence was sufficiently strong to be distinguishable from autofluorescence in the intestinal tube or spleen under a fluorescence dissecting microscope (Figure 1(C)). In PanIN region, the full spectrum of PanIN was observed in 68.7% and 36.4% of the total pancreatic area at 24 and 36 weeks after 4-OHT treatment, respectively (Figure 1(D-G)). Taken together, these data substantiate KRAS G12V mutation as an initiating event in the PanIN/PDAC sequence in pancreatic cancer.

Histological and immunohistochemical profiles of PanIN regions
To determine whether these experimental tumors resembled human PanIN, their histological and immunohistochemical profiles were examined. In the whole abnormal area, several grades of differentiation ranging from PanIN-1 to PanIN-3 were noted (Figure 2(A-D)). In PanIN-3 lesions, significant nuclear atypia and complete loss of polarity were observed, such that true cribriforming; i.e. budding off small clusters of epithelial cells into the lumen, obscured the distinction between luminal and basal boundaries (Figure 2(A,B)). The development of papillary, micropapillary or pseudostratified ductal lesions without nuclear atypia identified these as PanIN-1B (Figure 2(C,D)).
To identify the characteristics of PanIN regions, Alcian blue staining, immunohistochemistry against CK18, and immunofluorescence against cadherin1 were performed. Consistent with human PanIN, an abundant mucin content of PanIN was demonstrated by intense Alcian blue staining as a marker of mucin accumulation (Figure 2(A1-D1)). PanIN regions also showed highly positive staining with ductal specific marker, CK-18, suggesting their characteristics as ductal cells (Figure 2 (A2-D2)). Furthermore, intense cadherin1 staining was observed in the membrane of PanIN regions, confirming their identity as a characteristic pancreatic cancer precursor lesions (Figure 2(A3-D3)). DCLK1, a marker for gastric tuft cells, is also observed in the surface epithelium of PanIN lesions and in the intervening stroma in human pancreatic adenocarcinoma (Sureban et al. 2011). DCLK1 staining was performed on the PanIN in addition to pancreas region from groups without 4-OHT treatment as a control. In the control group, staining was observed in tuft cells of the GI tract but not in the pancreas ( Figure  3(A)). However, in the PanIN and the surrounding abnormal regions, DCLK1 staining was observed (Figure 3(C,F,  G)). The proliferative index of PanIN was assessed by expression of PCNA. In the group without 4-OHT treatment, the gastrointestinal epithelium underwent rapid cell turnover and the intestinal stem cells situated in the crypt of the fingerlike intestinal villi showed highly positive PCNA staining (Figure 3(A,B)). However, the percentage of PCNA positive cells in the pancreas was low, measuring 0.50% (Figure 3(E)). Conversely, in PanIN and surrounding abnormal regions, the basal level of PCNA expression was high, measuring 4.01% (Figure 3 (C-E)).
In addition, to determine the status of downstream signaling pathways known to be activated by oncogenic KRAS, we assessed levels of phospho-AKT using immunohistochemistry. In contrast to the infrequent AKT phosphorylation observed in groups without 4-OHT treatment (Figure 4(A,B)), PanIN regions showed widespread labeling for phospho-AKT (Figure 4(C,D)).

Discussion
In this study, we established the first KRAS-initiated pancreatic cancer model that closely recapitulates human PanIN. To reach the breakthrough in establishing zebrafish PanIN/PDAC model, we used a combination of CRE/Lox and GAL4/UAS systems. CRE/Lox system temporarily activated the GAL4 transactivator in the pancreatic progenitor region, whereas the GAL4/UAS system amplified transcription of the target gene. Thus, oncogenic KRAS G12V expression could be enhanced within the presumed ptf1a region of the pancreas. This strategy induced a KRAS-initiated pancreatic cancer model in zebrafish as evidenced by the development of the whole spectrum of PanIN. The histological and immunohistochemical profiles of these PanIN bore striking resemblance to human PanIN. Thus, we propose that the established KRAS-initiated pancreatic cancer model, generated by use of the combination of CRE/Lox and GAL4/UAS systems, now provides a preclinical platform to test various strategies for pancreatic cancer and finally improve outcomes for pancreatic cancer patients.
However, we also admit that our established KRASmediated model also needs some modifications, as this combined system induced PanIN at relatively a low rate (11.1%). As the lower incidence may impede its usage for drug screening and development, we propose several approaches to increase the induction frequency in our zebrafish model. First, a p53 mutant background may increase the PanIN incidence in the zebrafish model. p53 mutation occurs in approximately 50-75% of human PDAC and plays a critical role in cell cycle regulation by inhibiting growth arrest (Ryan et al. 2001). The significance of p53 mutation in pancreatic cancer is highlighted by the finding that the endogenous expression of both oncogenic KRAS and p53 mutant in the mouse pancreas dramatically shortened median survival months as compared with KRAS activation alone (Hingorani et al. 2005). Second, the coactivation of  Notch and KRAS may increase the frequency of PanIN formation in the zebrafish model. Notch activation is known to confer sensitivity to oncogenic KRAS by inducing the phenotypic plasticity of adult exocrine cells (Miyamoto et al. 2003;Siveke et al. 2008). Support for this finding is evident from the observation that coactivation of Notch and oncogenic KRAS dramatically increased PanIN formation compared with KRAS activation alone in a mouse model (De La O et al. 2008). Third, cerulean treatment may increase the frequency of PanIN in the zebrafish model, as ceruline-induced acute or chronic pancreatitis accelerated oncogenic KRAS-induced PanIN/PDAC formation (Carriere et al. 2007;Guerra et al. 2007;Friedlander et al. 2009;Morris et al. 2010). Furthermore, concurrent with animal model data, patients with pancreatitis show a higher risk of developing PanIN/PDAC (Raimondi et al. 2010). Taken together, with the proposed approaches, the induction frequency of PanIN formation is warranted to be increased to be used as a platform for the functional annotation of somatic mutations identified in pancreatic cancer genomes.
The zebrafish has emerged as an excellent model organism in the study of cancer biology over the last several decades. Traditionally, the focus of zebrafish research was on developmental biology because of the clear advantages afforded by this organism, such as the large brood size, transparent embryos, ex utero development of the embryo, and short life cycle. The long history of research in the developmental biology of the zebrafish has led to the identification of diseases similar to those in humans, especially cancer, with a closely histopathological and molecular similarity being observed between human and zebrafish tumors (Amatruda et al. 2002). Furthermore, zebrafish models are considered to be well suited for preclinical highthroughput drug screening, as the optical transparency of the zebrafish embryo makes it possible to monitor the effect of the drug on tumors with regard to their initiation, progression, and metastasis (Beis and Stainier 2006;Liu et al. 2008). Until recently, the preferred drugs for use against pancreatic cancer have been capecitabine and oxaliplatin (Bullock et al. 2017). However, the combination of chemotherapy with capecitabine and oxaliplatin induces duodenal ulcer bleeding and deterioration of general medical condition including anorexia and neutropenia (Chung et al. 2017). Thus, there is a need for better drugs that can be used alone or in combination with capetiabine/oxaliplatin. We propose that our KRAS-induced-pancreatic model may serve as a useful tool for preclinical high-throughput drug screening against pancreatic cancer. In particular, our model affords the ability to monitor the severity of dedifferentiation or transdifferentiation in oncogenic KRAS-induced-pancreatic progenitor cells, providing an avenue for in vivo testing of drug efficacy.
In summary, we successfully developed a novel system combining CRE/Lox technology with the GAL4/ UAS system to establish an oncogenic KRAS-initiated pancreatic cancer model. Our novel system demonstrated that KRAS G12V -responsive pancreatic progenitor cells could induce PanIN, constituting a precursor region of PDAC. This zebrafish cancer model thus provides an experimental and preclinical model system to investigate the basic biology of pancreatic cancer and identify potential therapeutic targets.

Author contributions
JTP and SDL conceived of and designed the experiments. JTP performed the experiments. JTP and SDL wrote and edited the paper.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by an Incheon National University research grant .