Truncation of the tumour suppressor adenomatous polyposis coli (APC) constitutively activates the Wnt/β‐catenin signalling pathway. This event constitutes the primary transforming event in sporadic colorectal cancer in humans. Moreover, humans or mice carrying germline truncating mutations in APC develop large numbers of intestinal adenomas. Here, we report that zebrafish that are heterozygous for a truncating APC mutation spontaneously develop intestinal, hepatic and pancreatic neoplasias that are highly proliferative, accumulate β‐catenin and express Wnt target genes. Treatment with the chemical carcinogen 7,12‐dimethylbenz[a]anthracene accelerates the induction of these lesions. These observations establish apc‐mutant zebrafish as a bona fide model for the study of digestive tract cancer.
APC (adenomatous polyposis coli) was originally identified as the gene mutated in the familial adenomatous polyposis (FAP) syndrome in humans (Groden et al, 1991; Kinzler et al, 1991). Carriers of germline truncating mutations in APC develop multiple colorectal adenomas following somatic inactivation of the remaining allele (Kinzler & Vogelstein, 1996). Similarly, mice that are heterozygous for a truncating mutation in APC develop multiple tumours in the small intestine (Su et al, 1992). APC is a key inhibitor of the Wnt/β‐catenin signalling pathway (Bienz & Clevers, 2000). It is an essential component of the Axin‐containing destruction complex that phosphorylates β‐catenin, thus tagging it for ubiquitination and degradation by the proteasome. In the presence of a Wnt ligand, β‐catenin is stabilized and accumulates in the nucleus, where it associates with T‐cell factor (TCF) family transcription factors and activates transcription of target genes. In sporadic cases of colorectal cancer, APC mutations cluster in a region proximal to the central Axin‐binding motifs, termed the mutation cluster region (MCR). Truncations at or proximal to the MCR abolish the inhibitory capacity of APC, leading to accumulation of β‐catenin and constitutive activity of the Wnt pathway (Korinek et al, 1997). A small percentage of sporadic colorectal cancers without mutated APC turn out to have mutations in the APC‐binding domain of β‐catenin that has the same effect (Morin et al, 1997).
The zebrafish provides a powerful animal model for genetic analysis of vertebrate embryogenesis, organ development and disease. Zebrafish is unique among vertebrates in its capacity for ‘forward’ (phenotype‐based) genetic analysis owing to a short generation time, large progeny size and readily accessible transparent embryos. Large‐scale forward genetic screens have identified hundreds of mutants with defects in embryogenesis (Driever et al, 1996). Many of these mutants have been proved to be excellent models for a variety of human diseases (Dooley & Zon, 2000). Cancer can occur in the wild in fish (Walter & Kazianis, 2001), and experimental carcinogenesis studies have demonstrated that fish develop tumours in virtually all organs (Spitsbergen et al, 2000). Most importantly, the histopathology of such neoplasia often resembles that of human tumours (Amatruda et al, 2002; Stern & Zon, 2003).
To explore the zebrafish as a model for β‐catenin‐driven carcinogenesis, we have studied adult heterozygotes carrying a previously identified mutation in the APC gene (Hurlstone et al, 2003). It is noteworthy that homozygous apc mutants show a pleiotropic, embryonic lethal phenotype that included a cardiac valve abnormality (Hurlstone et al, 2003). The truncating mutation (apcMCR) maps in the MCR, and is similar to mutations found in FAP patients. apc/+ fish were killed and assessed by histology for the presence of tumours. As a marker for proliferation, we used proliferating‐cell nuclear antigen (PCNA). Elevated levels of cellular β‐catenin indicated clonal loss of APC function.
We found that aged apc/+ fish (>15 months, n=34) developed spontaneous tumours, primarily in the liver (6 out of 34) and intestine (4 out of 34), whereas only one wild‐type (wt) sibling (n=33) had developed a neoplastic lesion at this age, which located in the liver.
The organization and self‐renewal of the epithelium of adult fish intestine resembles that of neonatal mammals, with proliferation being restricted to the intervillus pockets. Differentiation occurs along the crypt–villus axis (Crosnier et al, 2005; Wallace et al, 2005). With the exception of Paneth cells, fish possess all cell types present in the mammalian intestine. In the wt adult zebrafish intestine, low levels of β‐catenin immunoreactivity were detected in the epithelial cells of intestinal villi, largely restricted to the intercellular junctions (Fig 1A). Proliferative PCNA+ cells were restricted to the intervillus pockets (Fig 1D). In 15‐month‐old apc/+ fish, large outgrowths resembled mammalian polyps. The intestinal architecture was disorganized and in contrast to the ordered periodic arrangement of villi in the wt intestine, large structures with ramifications of the villi, often embedded in fibrovascular stroma, were observed (Fig 1B,C). These lesions were classified as adenomatous polyps, portraying pseudostratification of nuclei, loss of goblet cells and increased nuclear‐to‐cytoplasmic ratio consistent with dysplastic epithelium. Most of the cells in these intestinal adenomas had accumulated high levels of β‐catenin in the cytoplasm and nucleus (Fig 1B,C) and were proliferating as shown by PCNA staining (Fig 1E,F).
The adult zebrafish liver abuts the intestine in several locations, instead of being a discrete organ as in mammals. Although unlobulated, the histology of the zebrafish liver is otherwise similar to that of the mammalian liver and consists of differentiated hepatocytes, bile ducts, portal tracts and vessels (Fig 2A). In the wt liver, β‐catenin labels the membranes of the hepatocytes (Fig 2A) and there is minimal proliferation, as indicated by labelling for PCNA (Fig 2D). In 6 out of 34 apc/+ fish analysed at this age, we observed hepatic neoplasias, which were often associated with foci of hepatocellular alteration, clusters of hepatocytes with a more eosinophilic and a more vacuolated cytoplasm owing to glycogen accumulation (Fig 2C). In these lesions, we observed changes in cell morphology, such as condensed chromatin, prominent nucleoli and increased occurrence of apoptotic bodies (Fig 2F). These distinct features are reminiscent of hepatoblastoma in human infants. However, as they occurred in aged fish, they did not formally adhere to the classification criteria of hepatoblastomas. The hepatic lesions were circumscribed and maintained more of the original liver architecture and were tentatively classified as hepatic adenomas. The hepatic adenomas showed nuclear staining for β‐catenin, the hallmark of activated Wnt signalling (Fig 2B), and a high degree of proliferation (Fig 2E).
To further characterize the spontaneous intestinal and hepatic lesions in apc/+ fish, we assessed expression of the TCF4 target genes cmyc (He et al, 1998) and axin2 (Jho et al, 2002) by in situ hybridization. In the adult wt fish, cmyc expression is confined to the intervillus pockets of the intestine, and is undetectable in the liver (supplementary Fig 1Aref> online). axin2 expression is barely detectable in the adult fish intestine, and is undetectable in the liver (supplementary Fig 1Bref> online). Most of the cells comprising the intestinal adenomas were expressing high levels of cmyc (Fig 3A) and ectopic axin2 (Fig 3B). In the hepatic adenomas, high levels of cmyc and axin2 expression were found in scattered cells (Fig 3D,E) that were also PCNA positive (Fig 3F). As expected, the intestinal adenomas were negative for the differentiation marker intestinal fatty acid‐binding protein (ifabp; Fig 3C). Indeed, ifabp expression is downregulated on activation of Wnt signalling and in adenomas of min mice (van de Wetering et al, 2002). We next assessed the status of the wt APC allele in the lesions. The genomic region encompassing the MCR was amplified and sequenced from adenoma tissue along with adjacent normal tissue (supplementary informationref> online). All samples analysed were heterozygous for apc, indicating that in these adenomas there was no loss of the MCR in the wt allele and also that no other mutations were found in this genomic fragment.
We then sensitized wt and apc/+ siblings to tumour development by treating them with the carcinogen 7,12‐dimethylbenz[a]anthracene (DMBA). DMBA induces point mutations and has been used previously in carcinogenesis experiments in mice and fish (Spitsbergen et al, 2000). Fish from an outcross of apc/+ with wt were treated at 3 weeks of age with 5 p.p.m. DMBA. A total of 82 fish were killed at 6 months of age and a further 58 fish were killed at 14 months of age. We found that DMBA‐challenged apc/+ fish developed intestinal, hepatic and pancreatic neoplasms (Table 1). In 35.3% of DMBA‐treated apc/+ fish analysed at 6 months, we observed small intestinal adenomas encompassing two or three intervillus pockets. Most of the cells in these lesions including the associated villi and surface epithelium contained high levels of β‐catenin and were proliferative (Fig 4A,B). In some cases, we observed structures resembling single overgrown crypts in which cells were piling up in a disorganized fashion, contained high levels of β‐catenin and were proliferative (supplementary Fig 2ref> online). These structures resembled dysplastic aberrant crypt foci, premalignant lesions also found in human FAP patients. In the same experiment, 10.4% of wt sibling fish showed similar intestinal lesions. In 44.1% of treated apc/+ fish assessed at 6 months, we observed hepatic neoplasias as compared with one wt fish at that age (2%). The neoplasms contained foci of highly proliferative cells with increased levels of β‐catenin (Fig 4C,D). The histology was similar to that of the above‐described liver adenomas, which can be considered as premalignant lesions (Spitsbergen et al, 2000). When we analysed older animals (14 months), we observed that a greater percentage of these apc/+ fish showed neoplasms, which were unsurprisingly also larger and more dysplastic (supplementary Fig 2Cref> online). A total of 58.3% of DMBA‐treated apc/+ fish showed intestinal neoplasms compared with 20.5% of DMBA‐treated wt fish. Liver lesions occurred in 70.8% of apc/+ fish, whereas only 20.5% of wt fish analysed at this age (n=34) showed β‐catenin‐positive hepatic adenomas. In DMBA‐treated livers, we also observed bile duct neoplasias, characterized by abnormal proliferation of the bile ductules and accumulation of β‐catenin (supplementary Fig 3ref> online). We observed these bile duct neoplasms in 14.7% of 6‐month‐old apc/+ fish compared with 4% of wt fish analysed at this age. A total of 29% of 14‐month‐old apc/+ fish showed bile duct neoplasms, whereas 8.8% of wt fish showed this tumour type. Individual DMBA‐treated apc/+ fish typically showed neoplasms in several digestive tract organs, whereas neoplasias were almost always confined to one organ in wt siblings.
Pancreatic neoplasias were also observed in DMBA‐treated apc/+ fish. In the adult wt pancreas, β‐catenin levels were low, and very limited proliferation was observed (Fig 5A,B). In 35.3% of challenged apc/+ fish analysed at 6 months, we observed large masses of acinar cells showing high levels of β‐catenin and excessive proliferation (Fig 5C,D). These tumours invariably involved the exocrine pancreas and never endocrine pancreas components. Tumours led to an overall expansion of the organ. Tumour cells showed dysplastic features and had often lost their secretory granules, resembling acinar cell adenomas rather than pancreatic duct lesions. Only 1 out of 48 (2%) DMBA‐treated wt fish analysed at 6 months showed a pancreatic tumour.
The current data establish that zebrafish APC, like its mammalian counterpart, is a bona fide tumour suppressor. Moreover, heterozygous loss of APC in the zebrafish appears to phenocopy the cancer phenotype in mammals: (i) the morphology and histopathological features of the neoplasias that we observed resemble those found in humans (this study; Amatruda et al, 2002) and (ii) the spectrum of organs that are affected by neoplastic lesions in APC‐mutant fish overlaps with that observed in FAP patients and min mice. In FAP patients, there is a high increase in the likelihood of developing liver and pancreatic tumours in addition to the intestinal tumours (Giardiello et al, 1991, 1993).
We did not detect somatic loss of the wt APC allele in the lesions we analysed, similar to the situation in FAP patients (Nagase & Nakamura, 1993). Nevertheless, our results show that APC function is lost in the cells comprising the adenomas, as documented by the accumulation of β‐catenin, the lack of differentiation and the upregulation of Wnt signalling target genes. We assumed that mutations could have occurred elsewhere in the large apc gene, or that the wt APC allele was silenced by epigenetic changes.
The great benefit of modelling cancer in zebrafish is the feasibility of performing genetic or chemical screens for modifiers of the cancer phenotype. This could ultimately lead to the identification of novel therapeutic targets. The early embryonic lethal phenotype of apc homozygous mutants (Hurlstone et al, 2003) could serve as a readout in screens to identify molecules or genes that could suppress or attenuate the cardiac phenotype. Once identified, these genetic or chemical modifiers could then be tested to assess their role in cancer incidence. Current fish models of cancer show unusual tissue tropism as with tp53 and ribosomal protein mutants that develop peripheral nerve sheath tumours (Amsterdam et al, 2004; Berghmans et al, 2005), or use transgenic overexpression of mammalian oncogenes to induce a cancer phenotype, such as activated BRAF in the case of melanoma in fish (Patton et al, 2005) or c‐Myc in the case of pancreatic neuroendocrine cancer (Yang et al, 2004) or T‐cell leukaemia (Langenau et al, 2003). The apc/+ fish could serve as a valuable model for screens, as the model maintains the sporadic nature of tumours, and closely follows the molecular pathogenesis of its mammalian counterpart.
Fish. The apcmcr allele has previously been described (Hurlstone et al, 2003). Fish were maintained in the AB or TL genetic backgrounds. Fish were cared for in accordance with institutional guidelines. Fish in the AB background were used for the DMBA studies. In the spontaneous cancer incidence studies, no difference in the incidence of neoplasias between the two backgrounds was observed. Genotyping of fish was carried out by PCR on isolated genomic DNA from tail clips, as described (Hurlstone et al, 2003).
7,12‐Dimethylbenz[a]anthracene treatment. wt and apc/+ fry at 3 weeks of age were immersed overnight in a solution containing 5 p.p.m. of DMBA (Sigma‐Aldrich‐Chemie BV, The Netherlands) in dimethyl sulphoxide. The next day, fry were rinsed several times in water and were returned to the aquarium. Fish were monitored regularly for signs of sickness or evidence of tumours. Fish were killed for analysis at 6 or 14 months after the DMBA treatment.
Supplementary information is available at EMBO reports online (http://www.nature.com/embor/journal/vaop/ncurrent/extref/7400638‐s1.pdf).
We thank H. Stern (Harvard Medical School, USA) for advice on DMBA treatment and S. Schulte‐Merker (Hubrecht Laboratory) for a critical reading of the manuscript. This work was supported by a Dutch Cancer Society (KWF) grant to H.J.C. and in part by an NWO‐Vidi grant to A‐P.G.H.
- Copyright © 2006 European Molecular Biology Organization