A genetic screen in zebrafish identifies the mutants vps18, nf2and foie gras as models of liver disease

Chronic liver disease is responsible for over 27,000 deaths/year in the USA(National Center for Health Statistics,2004). This number is predicted to rise because of the association of liver disease with obesity: as the numbers of obese individuals in the USA approaches one-third of the population, the incidence of fatty liver has been estimated to affect ∼25% of US citizens(Neuschwander-Tetri and Caldwell,2003). Several hepatobiliary diseases are hereditary or result from developmental defects, especially those presenting in children, such as metabolic disorders and bile stasis (cholestasis) resulting from paucity of bile ducts (Kelly and McKiernan,1998; Pall and Jonas,2005). However, with the exception of a few well-known cases,there is a wide chasm between the clinical and molecular understanding of these diseases.

The power of zebrafish to study vertebrate development is well appreciated,and many disease models have emerged from studies with this organism(Rubinstein, 2003). There are several advantages to using zebrafish embryos to develop models of liver diseases. First, in addition to the embryological and genetic benefits of zebrafish, the liver is not the site of embryonic hematopoiesis as it is in mammals, and therefore mutants in liver size or structure will not have the confounding phenotype of hematopoietic dysfunction. Second, the zebrafish embryo survives on yolk for the first 3-4 days of development, after which its digestive system is fully functional and the embryonic fish begins feeding on day 5. Thus, not only is embryogenesis rapid, but the development of a physiologically functional liver occurs within the span of a few days. This provides the opportunity to carry out an embryonic screen to identify mutants that are defective in either liver development or function, or both. Third,baring minor differences, the anatomy, function, organization and cellular composition of adult zebrafish and mammalian livers are virtually the same(Hinton and Couch, 1998; Rocha et al., 1994; Wallace and Pack, 2003), as is the histopathology of fatty liver (steatosis), cholestasis and neoplasia(Amatruda et al., 2002; Spitsbergen et al., 2000) (J. Glickman, personal communication), allowing direct comparison between zebrafish and mammalian liver disease processes. Indeed, recent work has demonstrated that genes that underlie Alagille Syndrome, a pediatric disorder that results in a paucity of intrahepatic bile ducts, among other defects,play an important role in zebrafish biliary development(Lorent et al., 2004). Finally, although the early stages of hepatogenesis are relatively well understood and are similar in mice and zebrafish(Duncan, 2003; Field et al., 2003; Ober et al., 2003), less is known in either system about the final stage of hepatogenesis – hepatic outgrowth. Therefore, the zebrafish is an excellent system in which to screen for embryonic mutants with large livers (hepatomegaly), for it holds the potential of uncovering both developmental and pathological processes that contribute to this phenotype.

Our laboratory has used insertional mutagenesis to generate over 400 lines of zebrafish bearing mutations in 315 genes that are essential for embryonic development, and the mutated gene has been cloned for each line(Amsterdam et al., 2004; Golling et al., 2002). We calculate that this represents ∼22% of the total number of zygotic genes that are essential, as determined genetically, between days 1 and 5 of development (Amsterdam et al.,2004).

We developed a tool to screen day 5 embryos for abnormalities in liver size. Seven mutants with hepatomegaly were identified out of the 297 lines screened. Although several of the mutants may be defective in the regulation of hepatic outgrowth, three of them [vps18, neurofibromatosis 2(nf2) and foie gras (fgr; flj12716l– Zebrafish Information Network)] demonstrate signs of hepatic pathology. Mutation in the vps18 gene (which is required for endosomal trafficking to acidic organelles) results in albinism and hepatomegaly associated with enlarged hepatocytes, malformation of the bile canaliculi (which is the site of bile secretion at the hepatocyte apical membrane) and biliary paucity. Whereas the pigmentation defect and hepatocyte enlargement can be attributed to the failure to traffic endosomes to the correct intracellular compartment, the canalicular defect probably reflects a defect in the formation of the hepatocyte apical membrane. The hepatobiliary phenotypes seen in this mutant resemble the hepatic signs of individuals with arthrogryposis-renal dysfunction-cholestasis (ARC) syndrome, which is caused by mutation of vps33B (Gissen et al., 2004), a gene that interacts with vps18(Kim et al., 2001; Peterson and Emr, 2001; Sriram et al., 2003). Second,mutation of nf2 results in hepatomegaly associated with choledochal cyst formation, and we hypothesize that this results from deregulated biliary cell proliferation. Third, fgr mutants develop severe steatosis and we propose this mutant as the first non-mammalian model of a fatty liver disease.

The collection of mutants used in this study were generated, maintained and bred as described (Amsterdam and Hopkins,1999; Amsterdam et al.,2004). At least five wild-type and five mutant embryos from clutches containing healthy embryos were fixed in formalin overnight at 4°C, washed with PBST (PBS + 0.1% Tween 20) and dehydrated in methanol.

The labeling protocol was modified from that developed by C. Semino(Phylonix Pharmaceuticals). Embryos were rehydrated through a graded series of methanol to PBST, bleached with 10% H₂O₂/0.5×SSC/0.5% formamide for 12 minutes and blocked with PBST/10% BSA for 1 hour at room temperature. Embryos were incubated with CY3-SA (Sigma; 1:500) in a dark chamber for 2 hours at room temperature, washed with PBST and stored in 80%glycerol.

Embryos were viewed on a Leica MZFLIII stereomicroscope equipped with a fluorescent attachment and scored for liver size, shape, number of lobes as well as for labeling and morphology of the gut. Mutants with hepatomegaly had a left lobe that was visually estimated as greater than ∼20% larger than that of their phenotypically wild-type siblings. In each case, the general morphological phenotype has been shown to be tightly linked to a single viral insertion (Golling et al.,2002), and the liver phenotype was always associated with the general morphological phenotype. The mutants were coded, scored blind and only decoded after the phenotype had been observed in at least four clutches. Images were obtained on a Zeiss Axioplan 2 using OpenLab software(Improvision, Lexington, MA).

Approximately 0.5-2.0 nl of the following morpholinos at the indicated concentrations were injected into one-cell embryos: vps18 (1 mM)ATTGATCCAGAATAGATGCCATTGC; nf2 (0.5 mM) TCAGACCCAAATTGACATAGTGAC; fgr (0.05 mM) GAGATCCCATTGCGCTGGACTCATG.

Day 5 wild-type and mutant embryos from each line were collected and RNA was extracted using the RNAEasy kit (Qiagen). Oligo dT primed cDNA from the RNA equivalent of two embryos was created using the SuperScript II RT Kit(Invitrogen). PCR reactions (30 μl) contained 0.25% of the cDNA reaction,1× buffer, 0.4 mM MgCl₂, 0.2 μM dNTPs, 0.5 μl Taq polymerase (Invitrogen) and 0.4 μM of each primer, and the reaction was carried out for 30 cycles. PCR products were run on a 1.2% agarose gel containing 1 μg/ml ethidium bromide. Primers sequences, 3′ to 5′: fgr-F CTTGCCCCATGAGGTATGAGCAC, fgr-R TGTTGAGCTGAGGGAGGACT; vps18-F CTGGAGGTTGAACGTGGTTT, vps18-R GCAGGAGCAAGAAGTGGAAC; nf2-F CAACCCCACAACAAGCTGAGC, nf2-R GAAGATCGGCTGTTTCCTCAGAG; Actin-F CATCAGCATGGCTTCTGCTCT, Actin R-GCAGTGTACAGAGACACCC.

Embryos were fixed in 4% paraformaldehyde for 4 hours at room temperature,washed and dehydrated as described above and embedded in histogel(Richard-Allen Scientific) and then in paraffin. Serial sections (4 μm)were cut and stained with Hematoxylin and Eosin, photographed on a Leica DMRB microscope mounted with a QImaging Retiga EXi digital camera and processed using Adobe Photoshop 5.0.

Embryos for electron microscopy were fixed over night in Karnovsky's fixative (0.1 M cacodylate/2.5% gluteraldehyde/2% formaldehyde/0.85 M CaCl₂ pH 7.4) and processed by the Renal Pathology Service at the Brigham and Woman's Hospital (Boston, MA).

Day 5 and 7 embryos were anesthetized and embedded in OCT Compound (Tissue Tek, Sakura Fintek), frozen on dry ice and 10 μm sections were cut. Sections were thawed and fixed for 10 minutes with acetone cooled to–20°C. Non-specific sites were blocked with 1% BSA in PBS containing 0.05% Triton X100 and was incubated with 2 hours at room temperature with a 1:100 dilution of monoclonal antibody to P-glycoprotein (Mdr1 and 3, clone C219; Signet laboratories, Dedham, MA). Sections were washed and incubated with Cy3-SA (1:250; Sigma) and Alexa 488 anti-mouse IgG (1:1000; Molecular probes, Eugene, OR) for 1 hour at room temperature. Images were acquired on a Zeiss LSM 510 confocal microscope and processed using Adobe Photoshop.

Day 7 embryos were bathed in 0.3 μg/ml PED6 (Molecular Probes) for 2 hours at 28°C and imaged on a Ziess Axioplan 2. Over 30 mutant, wild-type and morphant embryos were scored for incorporation of the dye into the gallbladder and for the size of the common bile duct.

Lysosomes and nuclei were counted in 5 thin (1 μm), Toluidine Blue stained Epon sections through the liver of day 5 vps18 mutant embryos and their phenotypically wild-type siblings.

For each line with hepatomegaly, the hepatocyte internuclear distance was determined from Hematoxylin and Eosin sections. Ten images at 1000×magnification from at least two wild-type and three mutant embryos from the same clutch were collected and the distance between adjacent cell nuclei was measured for 125-175 hepatocytes using Openlab. T-tests were performed on each wild-type-mutant pair; P-values <0.001 were considered to be significant.

Our laboratory has used viral insertional mutagenesis to generate a collection of over 400 zebrafish lines that are mutated in 315 recessive embryonic lethal genes. The gene bearing the mutagenic insert has been cloned for each line and we estimate that we have achieved ∼22% saturation(Amsterdam et al., 1999). As many human genetic diseases are attributed to mutation of one allele of an essential gene that, when homozygously deleted, results in embryonic or neonatal mortality, we envisioned our collection as a pool of potential models for human genetic diseases. Indeed, a recent screen of our collection for mutants that develop kidney cysts has revealed a number of genes that may be important in this disease in humans (Sun et al., 2004).

Hepatomegaly is one of the most common and obvious signs of liver disease. Therefore, hepatomegaly in day 5 zebrafish embryos could be a sign of liver pathology. Alternatively, hepatomegaly could develop from a deregulated liver growth during hepatic outgrowth, and such mutants would be valuable for understanding this final phase of liver development.

We developed a technique to specifically label the liver so as to carry out a large scale screen using fixed material. As biotin serves as a cofactor for a number of liver enzymes (Moss and Lane,1971), it is found at high levels in hepatocytes. We found that streptavidin conjugated to the fluorophore CY3 (CY3-SA) labels the liver in day 5 embryos (Fig. 1A,B). Additionally, CY3-SA labels the intestinal epithelia and yolk because of a high concentration of biotin in these tissues(Fig. 1A,B).

We used CY3-SA to screen nearly all of the lines of our collection to identify those which develop hepatomegaly on day 5 of development. Out of the 297 lines that were available to be screened, seven mutants with hepatomegaly were identified (Fig. 1B; Table 1), representing 2.4% of all mutants screened. Given that our collection represents ∼22% of all embryonic essential genes, and that this screen covered 94% of our collection(i.e. ∼21% of all embryonic essential genes), we estimate that a total of∼33 of the essential genes could result in this phenotype. The genes underlying hepatomegaly each appear to function in different cellular processes, although two (fgr and pté) have no defined cellular or biochemical function (i.e. `novel'; Table 1).

Hepatomegaly in each mutant was confirmed by in situ hybridization with the liver-specific marker, fatty acid binding protein 1 (K.C.S., unpublished)(Her et al., 2003). Morphological and histological analysis did not reveal organomegaly in any other tissue (K.C.S., unpublished). The yolk is consumed by day 5 in most wild-type embryos; however, we observe that most of the mutant embryos in our collection, including the embryos identified by this screen, do not fully use the yolk (Fig. 1B). CY3-SA also labels the intestinal epithelia and we used this to score gut morphology for each mutant. The gut develops normally in sox9a, vps18 and nf2, is slightly underdeveloped in pté, thoc2 andγ -tub, and is severely abnormal in fgr.

Diagnosis of liver disease relies upon clinical presentation, biochemical measurement of liver function and assessment of the pathology from a liver biopsy. We used histological analysis to identify signs of pathology in each mutant, and to assess the hepatic architecture, cell composition and cell size. The cellular phenotype of each mutant is described in Table 1. Four mutants(vps18, fgr, γ-tub and thoc2) were identified as having enlarged hepatocytes, which was confirmed by a quantitative measure of cell size (Fig. 2). Two mutants (vps18 and nf2) contained large spaces in the extracellular hepatic matrix (Fig. 2A; see Fig. 5). Further analysis suggested that this is associated with biliary defects in both cases (see below). In two mutants (sox9a and pté), the hepatocytes are the same size as wild type(Fig. 2B) and there is no evidence of expansion of the extracellular space; in these cases, hepatomegaly may develop due to unregulated proliferation of hepatocytes. Signs of hepatobiliary pathology were detected in three mutants (vps18, nf2and fgr), and we further characterized each of these.

Mutants from hi^2499A contain an insertion in the vps18 gene following nucleotide 2236(Fig. 3A). This insertion results in an abrogation of the vps18 message in mutant embryos(Fig. 3B). vps18message is present in oocytes (i.e. immature and unfertilized eggs) and in embryos on days 1-5 of development (K.C.S., unpublished), and only the zygotically transcribed message is affected in the hi^2499A mutants. Injecting high concentrations of a morpholino directed against the start site of the vps18 gene to knock-down the translation of the maternal message results in pigmentation defects(Fig. 3A) and, in some embryos,causes global developmental defects early in development (K.C.S.,unpublished). Lower concentrations of the morpholino do not interfere significantly with pigmentation or development, but do cause hepatomegaly(Fig. 4A). These data indicate that the pigmentation defects and hepatomegaly observed in mutants from hi^2499A can be attributed to the loss of vps18 function. Thus, we hereafter refer to the hi^2499A mutants as vps18.

vps18 is a class C vacuolar protein sorting gene(Raymond et al., 1992). The products of the class C genes (vps11, vps16, vps18 and vps33) direct the targeting, docking and SNARE-mediated fusion of vesicles to the yeast vacuole and animal lysosome(Kim et al., 2001; Peterson and Emr, 2001; Poupon et al., 2003; Raymond et al., 1992; Sato et al., 2000; Srivastava et al., 2000). We carried out a detailed analysis to determine whether this function of the vps18 gene could account for the two prominent phenotypes displayed by vps18 mutants: albinism and hepatomegaly.

vps18 is well conserved and has an orthologs in every eukaryotic species examined (Fig. 3C). There is 66% identity between the human and zebrafish, and 33% between zebrafish and Drosophila proteins(Fig. 3D), suggesting that the function of vps18 is conserved between species. Abnormalities in eye pigmentation in the Drosophila vps18 mutant, deep orange(dor) (Puckett and Petty,1980; Sevrioukov et al.,1999) are due to a defect in endosomal trafficking and fusion with the pigment granule (Sevrioukov et al.,1999; Sriram et al.,2003). Although dense pigment granules are seen in the pigmented epithelia of the epidermis and retina of wild-type zebrafish embryos(Fig. 4B,D), vps18mutant embryos have only a few irregularly distributed pigment granules in these tissues (Fig. 4C-E). This phenotype resembles the pigmentation defect observed in dor mutants,suggesting that the function of vps18 in pigment granule maturation,which is analogous to lysosome biogenesis, is conserved in these two distantly related animals.

Given that loss of vps18 function disrupts the trafficking of late endosomes in yeast and mammals (Huizing et al., 2001; Poupon et al.,2003; Srivastava et al.,2000) we examined vps18 mutant livers to determine whether this same function could account for hepatomegaly that is observed in vps18 mutants and morphants (Fig. 5A).

Livers in day 5 wild-type embryos consist primarily of hepatocytes, which have a lacey, eosinophilic cytoplasm (Fig. 2A; Fig. 5B). Although sinusoids can be easily identified, bile ducts are never observed in embryonic livers using standard histological methods. Hepatocytes in vps18 mutant embryos are enlarged(Fig. 2) and contain large cytoplasmic structures resembling vesicles or vacuoles(Fig. 5D). There is less eosinophic material in the vps18 mutant hepatocytes compared with wild type, and this may reflect the decreased levels of stored glycogen in the vps18 mutant hepatocytes (K.C.S., unpublished).

We used transmission electron microscopy (TEM) to determine the nature of the cytoplasmic structures seen in vps18 mutant hepatocytes(Fig. 5E,F). Hepatocytes from wild-type day 5 embryos contain copious glycogen, which causes the cytoplasm to have a homogeneous, grainy appearance. By contrast, little glycogen is seen in the vps18 mutant hepatocytes. Instead, large, membrane-bound structures containing proteinacious material and debris were observed(Fig. 5F). These are similar to the aberrant structures seen in yeast and Drosophila cells that lack vps18 (Sevrioukov et al.,1999; Sriram et al.,2003; Srivastava et al.,2000). A failure to deliver endosomal cargo to the lysosome prevents lysosome biogenesis; we found that vps18 mutant hepatocytes have one-quarter the number of lysosomes of wild type(Fig. 4G). Taken together,these data suggest that vps18 acts in zebrafish to target endosomes to the pigment granule in melanocytes and to the lysosome in hepatocytes. We conclude that the failure of endosomal-lysosomal targeting causes transport intermediates to build up in the hepatocytes, resulting in hepatocyte enlargement and hepatomegaly.

ARC syndrome is an autosomal recessive disease that affects the liver,kidneys, platelets and neurogenic muscular function(Eastham et al., 2001). Typical hepatic defects include hepatomegaly, intrahepatic biliary paucity and cholestasis associated with mislocalization of canalicular markers(Gissen et al., 2004; Horslen et al., 1994). Recently, Gissen et al. have shown that this disease is due to mutation in another class C vps gene, vps33B(Gissen et al., 2004). Given that Vps18 and Vps33B function in the same complex(Huizing et al., 2001; Kim et al., 2001; Peterson and Emr, 2001; Poupon et al., 2003; Raymond et al., 1992; Subramanian et al., 2004), we asked whether biliary defects similar to ARC syndrome occur in vps18zebrafish mutants.

vps18 mutant livers contain large spaces in between hepatocytes(Fig. 2A). The spaces are devoid of nucleated red blood cells and were not lined by endothelial cells. We concluded that they were not part of the hepatic microvasculature(sinusoids). In order to determine whether these spaces reflected a defect in the intrahepatic biliary system, we undertook a detailed histological,immunological and ultrastructural examination of the embryonic biliary tree.

Bile is secreted at the hepatocyte apical membrane in a specialized structure called the bile canaliculi(Ujhazy et al., 2001). Bile transporters, such as MDR1, are specifically localized to the canaliculi in mammals (Trauner et al., 1997; Trauner and Boyer, 2003) and in fish (Hemmer et al., 1995; Lorent et al., 2004). In day 5 wild-type zebrafish embryos, MDR1 is localized exclusively to the tube-shaped canaliculi (Fig. 6B,C). By contrast, MDR1 localization in vps18 mutant livers is punctate(Fig. 6E,F) and the canaliculi are large and round (Fig. 6F,arrows). Importantly, we found MDR1 labeling in the cytoplasm of some vps18 hepatocytes (box in Fig. 6F), suggesting that it is not trafficked properly to the apical canalicular domain.

In mammals, bile flows from bile canaliculi to bile ducts through the canal of Hering. Teleosts do not appear to have a canal of Hering, but instead the canaliculi empty directly into the lumen of preductules, which are formed by biliary pre-ductal epithelial cells (PDEC), a specialized cell of the intrahepatic biliary tree in teleosts(Hinton and Pool, 1976; Rocha et al., 1994). PDECs form the lumen of the pre-ductule by wrapping around themselves, analogous to capillary formation by endothelial cells(Hinton and Couch, 1998; Hinton and Pool, 1976; Rocha et al., 1994). Although PDECs are reported to express cytokeratin 19(Lorent et al., 2004; Matthews et al., 2004), a classic marker of biliary epithelial cells, they are not columnar, they lack a basal lamina (Rocha et al.,1994) and they appear less differentiated than the cholangiocytes that form the hepatic duct, gallbladder and common bile duct(Fig. 7G).

We could not identify PDECs by histological or immunological methods, but could readily identify them by TEM as small cells that contain a prominent,multi-lobed nucleus, and are sandwiched between hepatocytes(Fig. 6G). The bile canaliculi in day 5 zebrafish embryos are formed at the apical membrane of 1-4 hepatocytes and are flanked by tight junctions between adjacent hepatocytes(K.C.S., unpublished) and between hepatocytes and PDECs(Fig. 6I). We found that the entire intrahepatic biliary tree in day 5 zebrafish embryos comprises preductules that coalesce at the hepatic duct.

TEM was used to determine whether vps18 mutants have a defect in PDECs and the bile canaliculi. The number of PDECs in day 5 vps18mutant livers is drastically reduced, and those we could identify were shrunken, did not maintain tight junctions with the surrounding hepatocytes and their nuclei contained condensed DNA(Fig. 6H), suggestive of apoptosis. The space seen to surround PDECs in vps18 mutants may represent the large extracellular spaces observed on histological sections. As in ARC syndrome patients (P. Gissen, personal communication) the extrahepatic biliary tree is not affected in vps18 mutants (K.C.S.,unpublished).

Consistent with the abnormal MDR1 labeling seen in vps18 mutant livers, there are dramatic defects in the bile canaliculi. Canaliculi in wild-type zebrafish embryos are packed with regularly spaced and evenly shaped microvilli (Fig. 6I), and ultrastructurally appear very similar to canaliculi in mammals. Canaliculi in vps18 mutants have sparse, blunted microvilli and are distended and contain debris (Fig. 6C,D). This morphology is typically seen in individuals with cholestasis and is identical to those seen in sea lamprey which develop cholestasis during metamorphosis (Sidon and Youson,1983), suggesting cholestasis may also develop in vps18mutants.

Our analysis of vps18 mutants indicates that the pigmentation defects and hepatomegaly can be attributed to the well characterized function for this gene in endosomal-lysosomal trafficking. In addition, this study has identified a new role for vps18 in trafficking to the hepatocyte apical plasma membrane and formation of the bile canaliculi. Canalicular malformation (Gissen et al.,2004) and biliary paucity (P. Gissen, personal communication) is an abnormality shared between vps18 mutants in zebrafish and mutation of the vps33B gene in individuals with ARC syndrome. Taken together,these data are consistent with the hypothesis that class C VPS genes are required in trafficking to the bile canaliculi at the hepatocyte apical plasma membrane.

Choledochal cyst formation is a congenital disorder that is most often detected in childhood, although an increased incidence in adults has recently been reported (Soreide et al.,2004). Cyst classification is based on the site of the cyst and whether both intrahepatic and extrahepatic involvement is detected(Soreide et al., 2004). The etiology of cyst formation is not known. Although most cases are the result of developmental defects, a genetic component is suggested by the occurrence of some familial cases (Behrns et al.,1998; Iwama, 1998; Iwama et al., 1985; Iwata et al., 1998). Choledochal cysts carry a greatly elevated risk for developing cholangiocarcinoma, and although cyst excision diminishes this risk, it is not eliminated (Soreide et al.,2004). This suggests that that cholangiocytes in individuals with choledochal cysts may be predisposed to hyperproliferation and malignant transformation.

The Nf2 gene in mammals is a tumor suppressor(Lekanne Deprez et al., 1994; McClatchey et al., 1998; Ruttledge et al., 1994; Sanson et al., 1993). Mutants from hi³³³² contain a single viral insertion within the intron preceding the first coding exon of the nf2 gene(Fig. 7A). This insertion causes a complete abrogation of the nf2 message in mutant embryos(Fig. 7B). The nf2gene has been duplicated in zebrafish to create nf2a (subsequently called nf2) and nf2b genes, which are over 60% identical at the nucleotide level (A.A., unpublished). The nf2 gene is expressed in oocytes and in embryos through day 5 of development, while the nf2b message is detected only in oocytes (K.C.S., unpublished). Injecting embryos with a morpholino that is specific for the nf2message phenocopies the biliary phenotype observed in mutants from hi³³³² (see below). Thus, it is unlikely that nf2b significantly affects the phenotype of mutants bearing an insertion in the nf2 gene and we refer to the mutants from hi³³³² as nf2.

Mouse embryos homozygously deleted for the nf2 gene arrest before gastrulation (McClatchey et al.,1997). Our screen indicates that the nf2 gene is also essential for embryogenesis in zebrafish, although nf2 mutants have only modest morphological defects on day 5 of development, including a small forebrain and hepatomegaly (Fig. 1B). We attribute this difference to the presence of maternal nf2 message, as injecting a high concentration of a morpholino designed against the nf2 start site arrests embryos at the one- to two-cell stage.

There is no dramatic hepatocellular phenotype in nf2 mutant livers, and the size of the hepatocytes are equivalent to wild-type embryos(Fig. 2C, Fig. 7B-E). In some regions of the liver, however, there are large spaces between hepatocytes(Fig. 7D and Table 1). These spaces do not appear to be sinusoids, but may reflect a biliary defect.

The most striking phenotype of nf2 mutants is the dilated common bile duct (Fig. 7G-N). On histological sections, both the gallbladder and the common bile duct are enlarged compared with wild-type sections through the same plane(Fig. 7G-I), indicating the formation of a type Ic (solitary, cystic) choledochal cyst. In some nf2 mutant embryos, diverticuli are observed in the cystic ducts(Fig. 7I), indicating the formation of Type II (supraduodenal diverticuli) choledochal cysts. Both of these abnormalities are not accompanied by any obvious morphological or dysplastic change in the cholangiocytes, but there does appear to be an increase in the number of cells forming the duct.

The fluorphor-linked phospholipid, PED6, has been used in zebrafish to visualize the gut and gallbladder in live embryos(Farber et al., 2001). We used PED6 to examine the common hepatic duct in live embryos. All wild-type day 7 embryos incubated in PED6 demonstrate robust fluorescence in the gut and gallbladder, as do the majority of the nf2 mutants(Fig. 7J,K), although some of the mutants and nf2 morphants fail to transfer any PED6 to the gallbladder (i.e. `no label' in Fig. 7O) The common bile duct can barely be detected as a string of fluorescence connecting the gallbladder and the gut in wild-type embryos(Fig. 7M,O), while 80% of the nf2 mutant embryos have a markedly distended bile duct(Fig. 7N,O). Roughly half of the mutant embryos also form a diverticuli from the common bile duct(Fig. 7N), confirming the histological phenotype (Fig. 7I). Obstruction can result in biliary dilation; however, serial sectioning through the common bile duct nf2 embryos as well as pulse chase experiments with PED6 did not reveal any obstruction distal to the cyst(K.C.S., unpublished). Taken together, these histological and physiological data indicate that the nf2 gene in zebrafish is involved in bile ductogenesis, and that loss of nf2 function results the formation of type I and type II choledochal cysts.

Steatosis is a common cause of hepatomegaly(Neuschwander-Tetri and Caldwell,2003). Approximately 10% of individuals with steatosis that have no history of alcohol abuse progress to develop the severe fatty liver disease non-alcoholic steatohepatitis (NASH), characterized by hepatomegaly, deranged liver function and steatosis, resulting in hepatocyte death and inflammation that can develop into cirrhosis (Diehl,2001). It is not clear why only a subset of individuals progress from simple steatosis to NASH, but genetic predisposition may play an important role. Indeed, while steatosis and NASH are usually associated with obesity, there are well documented cases of familial NASH(Neuschwander-Tetri and Caldwell,2003; Struben et al.,2000; Willner et al.,2001), indicating a genetic component to this disease.

hi^1532B mutants develop massive hepatomegaly, a failure of gut development and abnormalities lower jaw and the fin morphology(Fig. 1). hi^1532B mutants contain an insertion in the intron between exons 11 and 12 of a novel gene which we named foie gras(fgr). The virus contains a 172 bp gene-trap cassette and in hi^1532B mutant embryos, the gene-trap is spliced in frame following bp 1287 of the fgr-coding sequence(Fig. 8A). This results in a frame shift that creates a stop codon immediately following the gene trap(Fig. 8A). Thus, mutant embryos from hi^1532B contain only transcript that encodes the allele with the gene trap (Fig. 8B, mutant band), while phenotypically wild-type siblings from the same clutch, of which two-thirds of these embryos are heterozygotes, have transcript encoding the wild-type allele as well as a small amount of the mutant allele (Fig. 8B). Injection of a morpholino designed against the start codon of the fgrmessage phenocopies the hi^1532B mutants (K.C.S.,unpublished), leading us to conclude that the viral insertion in this gene results in a loss of fgr function. Henceforth, we refer to mutants from hi^1532B as fgr.

Histological analysis of fgr mutant livers revealed enlarged hepatocytes which are filled with large, clear vesicles(Fig. 8C-D), suggestive of fat accumulation. Cells with fragmented nuclei and cell corpses, indicative of cell death, were frequently seen in fgr mutant livers(Fig. 8D), but never in wild type. Using the lipid stain, oil red O, we found a substantial amount of lipid accumulation in fgr mutant livers(Fig. 8E,F). Thus, some of the hallmark signs of NASH – hepatocytes enlargement, steatosis and an increase in hepatocyte death – were all observed in fgrmutants. The exception is the conspicuous absence of inflammation in fgr mutant livers, despite the marked cell death. We attribute this to the incomplete maturation of the zebrafish immune system at this stage of development. Nevertheless, mutation of the fgr gene, for which the function has not yet been determined in any organism, results in hepatomegaly associated with a phenotype that resembles NASH, and may serve as a non-mammalian model for studying this widespread and important disease.

We have undertaken a forward genetic screen in zebrafish to identify mutants which develop hepatomegaly – a common sign of many liver pathologies – in an effort to develop zebrafish models of liver diseases. This study has not only identified three such models, but also demonstrated important roles for several new genes in liver development,physiology and pathology. We found signs of hepatic pathology in three mutants, and propose vps18 as a model for the hepatobiliary defects observed in ARC syndrome, nf2 as a model for choledochal cysts and fgr as a model for fatty liver disease.

The role of vps18 in vesicle trafficking to the vacuole in yeast(Peterson and Emr, 2001; Sato et al., 2000; Srivastava et al., 2000), to the pigment granule and synapse in Drosophila(Narayanan et al., 2000; Sevrioukov et al., 1999; Shestopal et al., 1997) and to the lysosome in mammals (Huizing et al.,2001; Kim et al.,2001; Poupon et al.,2003) has been well defined. Although data from Drosophila (Counce,1956) and our morpholino experiments suggest an essential role for vps18 early in development, we believe the maternal contribution of RNA, protein and nutrients in zebrafish enables the vps18 mutants,and others (such as γ-tub) that bear mutation in a cell-essential gene, to survive the first few days of development.

We found that the two prominent morphological phenotypes of the vps18 mutant zebrafish embryo – pigmentation and hepatomegaly– could both be attributed to the well characterized role for the Vps18 protein in trafficking to the pigment granule or lysosome. It is likely that hepatomegaly develops in these embryos due to the accumulation of prelysosomal vesicles in the hepatocyte cytoplasm causing cell enlargement. Further analysis of this mutant, however, suggests that this protein also functions in trafficking to the hepatocyte apical membrane and formation of the canaliculi,and is required for development of the intrahepatic biliary tree. Alternatively, biliary paucity in these mutants could develop because of damage and death of the PDECs (i.e. biliary atresia), possibly as a result of cholestasis which may develop in these mutants. Biliary paucity and canalicular defects are also seen in individuals who suffer from ARC syndrome(P. Gissen, personal communication), which is due to homozygous mutation of the vps33B gene (Gissen et al.,2004). Given that the Vps18 and Vps33B proteins act as part of the same complex in yeast and in animals(Huizing et al., 2001; Kim et al., 2001; Peterson and Emr, 2001; Sato et al., 2000) in trafficking to the vacuole and lysosome, and that vesicle targeting and fusion with the apical membrane in hepatocytes involves a SNARE-dependent process that requires other VPS genes (Tuma and Hubbard, 2001,), our data support the hypothesis that vps18 and vps33B are also required for trafficking to the apical membrane.

Interestingly, morpholino knock down of vps33B message in zebrafish embryos reportedly results in cholestasis and biliary paucity (M. Pack, personal communication). Given that the phenotypes of vps18mutants, vps33B morphants and individuals with ARC syndrome are not identical [i.e. vps18 mutants do not have signs of arthrogryposis or renal dysfunction (K.C.S., unpublished)] and vps33B zebrafish morphants and individuals with ARC syndrome are normally pigmented (M. Pack and P. Gissen, personal communication), we propose that vps18 and vps33B may take on tissue specific roles or that functional redundancy exists in some tissues but not in others. Our data suggest,however, that the hepatic phenotype of vps18 mutants and individuals with ARC syndrome are very similar, and that this is due to the interruption of the same molecular complex involved in two trafficking pathways in hepatocytes. We thus propose the vps18 zebrafish mutant as a model for studying the hepatic signs of ARC syndrome.

The other mutant we identified to have hepatomegaly associated with a biliary defect bears an inactivating mutation in the nf2 gene. These mutants develop type Ic and type II choledochal cysts. Although there are signs of intrahepatic biliary pathology in nf2 mutants, unlike vps18, the PDECs in nf2 mutants are not atretic, but on the contrary, may be hyperproliferative (K.C.S., unpublished). Indeed, the fact that intrahepatic bile ducts are often hyperproliferative in individuals with extrahepatic choledochal cysts (Nambirajan et al., 2000) and choledochal cysts carry a predisposition to cholangiocarcinoma (Soreide et al.,2004) indicates a strong link between cyst formation and proliferation control. It is possible that in the absence of nf2,there may be unregulated proliferation of a bi-potential cell that can give rise to both hepatocytes and biliary cells, analogous to an oval cell(Newsome et al., 2004). This could account for both the increase in biliary cells and hepatomegaly observed in nf2 mutant zebrafish embryos, and is consistent with the finding that heterozygous deletion of nf2 in mice leads to the formation of hepatocellular carcinomas following loss of heterozygosity at the nf2locus (McClatchey et al.,1998) and with what is observed in a liver-specific deletion of nf2 in mice (M., Curto and A. McClatchey, personal communication).

The Nf2 protein merlin is a member of the ERM (ezrinradixin-moesin) family,which serve as membrane-cytoskeletal linkers. Merlin is thought to be required for contact-mediated inhibition of growth through stabilization of adherens junctions (Lallemand et al.,2003) and also through the inhibition of Pak1(Kissil et al., 2003). Our preliminary studies indicate that tight junctions are not disrupted in nf2 mutant livers (K.C.S., unpublished). It is interesting that using RNAi to knock down the levels of a C. elegans ERM gene, erm-1,results in the formation of cyst in every tubular epithelial organ in which the gene is normally expressed (Gobel et al., 2004), although junctions are not affected in this model. It will be of interest to examine the structure of the cytoskeleton, junctions and the proliferative index of biliary cells and hepatocytes in nf2mutant zebrafish.

Although ARC syndrome and choledochal cysts are relatively rare, steatosis is found in ∼25% of the population of the USA, and nearly 1-2% of US citizens have fatty liver disease(Neuschwander-Tetri and Caldwell,2003), which places it among the most common hepatic pathologies in the developed world (Clark et al.,2002; Neuschwander-Tetri and Caldwell, 2003). Although obesity and excessive alcohol intake are the greatest contributing factors to steatosis and NASH, the clustering of NASH within some families (Struben et al.,2000; Willner et al.,2001) and the numerous mouse models that develop NASH points to a considerable yet varied genetic component to this disease.

We therefore predicted that we might uncover a model of steatosis through a screen for hepatomegaly, but were surprised to find that fgr so closely resembles NASH, with the noted exception that there is no inflammation in this mutant. Moreover, as the well-conserved fgr gene has no identifiable domains or motifs, it represents a truly novel factor regulating fat accumulation. Multiple molecular pathways, including those controlled by SREBP1-c, PPARγ and insulin are important regulators of hepatic steatosis (Browning and Horton,2004), and it will be of interest to see whether fgrplays a role in any of these pathways.

In summary, by screening for zebrafish mutants with hepatomegaly, novel and physiologically relevant genes underlying embryonic hepatomegaly have been uncovered. We have identified three mutant lines that can serve as valuable models of diseases of lysosomal trafficking, choledochal cyst formation and fatty liver disease. Additionally, several mutants may also provide insight onto the processes that control liver growth and size during development.

This work would not have been possible without the expert assistance of our fish staff, especially S. Farrington and J. Davenport, and the histology core,especially A. Caron. We also thank H. Rennke and C. Ford for their excellence in electron microcopy. The bioinformatics support provided by C. Whittaker and the consultation with the talented pathologists R. Bronson, M. McLaughlin and J. Glickman were invaluable. We are indebted to C. Ukomadu, P. Gissen, A. McClatchy and members of the Hopkins' laboratory for helpful discussions, and to K. Haddix for help with statistics. We are particularly grateful for guidance from J. Lees. K.C.S. was supported by a Ruth L. Kirschstein fellowship from the NICHD (1F32HD042920-01).