The Role of ARF6 in Biliary Atresia

Background & Aims Altered extrahepatic bile ducts, gut, and cardiovascular anomalies constitute the variable phenotype of biliary atresia (BA). Methods To identify potential susceptibility loci, Caucasian children, normal (controls) and with BA (cases) at two US centers were compared at >550000 SNP loci. Systems biology analysis was carried out on the data. In order to validate a key gene identified in the analysis, biliary morphogenesis was evaluated in 2-5-day post-fertilization zebrafish embryos after morpholino-antisense oligonucleotide knockdown of the candidate gene ADP ribosylation factor-6 (ARF6, Mo-arf6). Results Among 39 and 24 cases at centers 1 and 2, respectively, and 1907 controls, which clustered together on principal component analysis, the SNPs rs3126184 and rs10140366 in a 3’ flanking enhancer region for ARF6 demonstrated higher minor allele frequencies (MAF) in each cohort, and 63 combined cases, compared with controls (0.286 vs. 0.131, P = 5.94x10-7, OR 2.66; 0.286 vs. 0.13, P = 5.57x10-7, OR 2.66). Significance was enhanced in 77 total cases, which included 14 additional BA genotyped at rs3126184 only (p = 1.58x10-2, OR = 2.66). Pathway analysis of the 1000 top-ranked SNPs in CHP cases revealed enrichment of genes for EGF regulators (p<1 x10-7), ERK/MAPK and CREB canonical pathways (p<1 x10-34), and functional networks for cellular development and proliferation (p<1 x10-45), further supporting the role of EGFR-ARF6 signaling in BA. In zebrafish embryos, Mo-arf6 injection resulted in a sparse intrahepatic biliary network, several biliary epithelial cell defects, and poor bile excretion to the gall bladder compared with uninjected embryos. Biliary defects were reproduced with the EGFR-blocker AG1478 alone or with Mo-arf6 at lower doses of each agent and rescued with arf6 mRNA. Conclusions The BA-associated SNPs identify a chromosome 14q21.3 susceptibility locus encompassing the ARF6 gene. arf6 knockdown in zebrafish implicates early biliary dysgenesis as a basis for BA, and also suggests a role for EGFR signaling in BA pathogenesis.


Introduction
Biliary atresia (BA), which is characterized by jaundice and loss of extrahepatic bile ducts (EHBD) at birth, affects 1:18000 Caucasian children and is fatal without surgical intervention [1]. The BA disease phenotype is broad and understanding the molecular and developmental basis of BA can improve diagnosis and selection of treatment. This task poses significant challenges, because BA presents shortly after birth suggesting an origin early during fetal development in the embryologic type of BA, or perinatally in the isolated, potentially infectious/ inflammatory type of BA. Either type precludes the elucidation of conclusive evidence. This evidence would ideally document disease progression in sequential fetal tissue. Phenotypic heterogeneity consisting of several associated extrahepatic anomalies further defies either a unified mechanistic hypothesis or a perfect experimental model, but suggests that multiple susceptibility loci may be involved. Extrahepatic manifestations can include gut or cardiovascular anomalies including laterality defects such as asplenia/polyspenia and heterotaxy to varying degrees [2,3]. Disease complexity is compounded further by histological features which can overlap with those of other cholestatic conditions such as neonatal cholestasis, ductal plate malformation, and Caroli's disease [2,4,5]. A common pathophysiologic finding is failure to excrete bile from the liver on nuclear imaging, mandating surgical drainage with portoenterostomy. This procedure fails in nearly half of all patients leading to bile stasis and cirrhosis. As a result, BA accounts for over a third of all pediatric liver transplants, worldwide [6]. Poorly developed intrahepatic bile ducts can also explain failure of portoenterostomy. Such a lesion cannot be identified with certainty amidst histological sequelae of biliary obstruction in most cases. These features include reactive intrahepatic bile duct proliferation, cholestasis, fibrosis and cirrhosis, which can distort liver architecture beyond recognition [7].
The role of environmental factors is underscored by the association of BA with rotavirus, reovirus, and cytomegalovirus infections, and the fact that these viruses induce comparable lesions experimentally. However, these models do not provide insight into the developmental basis of BA suggested by the presence of 'minor' extrahepatic anomalies even in the isolated variety of BA [8,9,10,11].
Toward generating novel unbiased hypotheses with evidence from diseased human subjects, two sets of genome-wide association studies (GWAS) have provided new directions. In two North American children with BA, Leyva-Vega et al. [12] observed a common 1.76 heterozygous deletion in chromosome 2q37. 3. The deletion was inherited from the father who did not have any liver disease in one of the two affected children, confirming that other loci and/or environmental factors were also necessary in BA pathogenesis. These findings were confirmed by Cui et al. in 6 of 61 children with BA from the same investigator group, in whom the deletion was confined to the GPC1 gene within the same locus [13]. A novel developmental model was used to generate supportive evidence, which consisted of impaired biliary network development after gpc1 knockdown with morpholino antisense oligonuleotide (MO) in zebrafish embryos. In 339 Chinese children with BA, Cheng et al. [14] confirmed the association between BA and an intergenic chromosome 10q24 SNP locus, rs17095355, which had been reported previously by Garcia-Barcelo et al. [15] from the same group. This SNP locus indicated the involvement of the neighboring ADD3 gene in BA, because of differential immunostaining of the ADD3 protein when diseased and normal liver tissue was compared. Recently, Tsai et al. [16] also found an association between BA and one of 333 SNPs in the chromosome 10q24 locus in 171 North American BA patients. However, the best associated SNP, rs7099604, was located in the first intron of ADD3 in this locus. Interestingly, knockdown of add3 in zebrafish also impaired biliary network formation providing supportive evidence for yet another candidate gene identified with GWAS.
Together, these observations reinforce the idea that multiple susceptibility loci along with environmental factors may be involved in BA, and suggest that replicating a GWAS finding may be challenging. Adding to this difficulty is the fact that the SNPs best associated with BA were located in different regions in the chromosome 10q24 susceptibility locus in Chinese and North American patients. Finally, unrecognized minor extrahepatic anomalies, which are associated even with the 'isolated' "non-syndromic" variety of BA, can introduce trait heterogeneity in any test cohort. Unrecognized disease trait heterogeneity is a common cause of failure to replicate an association [17]. These limitations in determining BA pathogenesis with GWAS studies are well served by the availability of a well-characterized zebrafish model of bile duct development. This model can provide supportive evidence for additional novel susceptibility genes also identified by GWAS.
To identify additional susceptibility loci we performed GWAS in Caucasian BA cases and controls who presented to the Children's Hospital of Pittsburgh (CHP) and the Center for Applied Genomics (CAG) at the Children's Hospital of Philadelphia. This analysis identified ARF6 as a leading candidate gene involved in bile duct development, based on SNP associations. To identify other ARF6-related pathways members, the top-ranked SNPs identified by GWAS were further evaluated with a systems biology analysis. Similar to previous GWAS studies, we used MO-mediated translation blockade to obtain supportive evidence for the role of arf6 in the early bile duct development in zebrafish.

Materials and Methods
Children with BA were enrolled at CHP after obtaining written consent from parents for the University of Pittsburgh IRB approved protocol number 0405628. At CAG, all subjects were enrolled after obtaining written consent from parents per Children's Hospital of Philadelphia IRB approval number 06-004886. Genotyping of DNA obtained from blood was performed with the Infinium HumanHap550K BeadChip (Illumina, San Diego, CA). SNP genotype frequencies were compared between cases and controls with a chi-square test statistic applied in Plink [18] for SNPs with at least 90% call rate and 1% minor allele frequency (MAF). Fourteen additional patients were genotyped with TaqMan 1 SNP Genotyping Assay (Life Technologies ID; C_11918520_10) for the candidate SNP; rs3126184. Taqman genotyping assay was not available for rs10140366. All zebrafish experiments were performed with approval from the Institutional Animal Care and Use Committee (IACUC) of the University of Pittsburgh, approval number 13102738.
ARF6 immunostaining of liver explants from children with BA Formalin fixed, paraffin embedded (FFPE) sections were stained on the Ventana Benchmark Ultra IHC staining platform using ARF6 polyclonal antibody (Abcam, Cambridge, MA). Antigen retrieval was performed using Ventana's proprietary ultraCC1 cell conditioning reagent and visualized with the uvDAB detection system. Sections were counterstained with hematoxylin.

SNP and Pathway Analysis
Among top-ranked SNP loci, we examined enrichment for other genes related to ARF6 and its associated EGFR pathway genes [19,20,21]. Network and enrichment analyses using Cytoscape, a popular network visualization and analysis tool, and Ingenuity Pathway Analysis (IPA) were performed [22,23]. First, a large human network was created in Cytoscape by integrating protein-protein interaction data from BIOGRID and STRING databases, transcription factor interaction from TRANSFAC, and pathway information from KEGG pathways [24,25,26,27]. From the STRING database that uses a scoring scheme between 0 and 1 based on predicted and experimentally validated protein-protein interactions, a 0.9 cutoff was used to extract highly probable protein-protein interaction. Then, a smaller network was created using only the first neighbors of genes that could be mapped from the top 1000 significant SNPs from the GWAS CHP cohort. For SNP-to-gene mapping, +/-20kb window was applied on the mapping file provided by the manufacturer for the Infinium HumanHap550K BeadChip (Illumina, San Diego, CA). A total of 2506 genes in the new local network was submitted to Ingenuity Pathway Analysis (IPA) for upstream regulator analysis and enrichment of canonical pathways and biological functional categories.

Chemical treatment
A 3.17 mM stock of the EGFR inhibitor AG1478 (Cayman Chemical) was prepared in 100% dimethyl sulfoxide (DMSO) and diluted to 1 or 4 μM with egg water. As a control, 0.1% DMSO solution in egg water was used.

Whole-mount in situ hybridization and immunostaining
Whole-mount in situ hybridization was performed as previously described [34]. Whole-mount immunostaining was performed as previously described [35], using the following antibodies

PED-6 and proliferation assay
PED-6 assay was performed by treating larvae with 0.3 μg/ml PED-6 (Life Technologies) for 3 hours as described previously [37]. Proliferation assay was performed by treating larvae with egg water containing 10 mM EdU (Life Technologies) and 1% DMSO. After one-hour EdU treatment, the larvae were harvested and processed for EdU labeling using the Click-it EdU Imaging kit (Life Technology).

Generation of GFP expression constructs for morpholino validation
The GFP coding region was amplified by PCR with a forward primer that contains a morpholino target sequence in front of GFP start codon and a reverse primer that contains its stop codon. The PCR products were cloned into the pGEM-T vector (Promega, Madison, WI) and their sequences were analyzed. Clones without any mutations were chosen and used for subcloning into the pCS2 + expression vector that contains the CMV promoter and the SV40 polyA site. Plasmids containing the final CMV: GFP constructs were used for injection.

qRT-PCR methods
Total RNA was extracted from pool of 100 livers of uninjected control and arf6-ATG MOinjected larvae at 4.5 dpf by using RNeasy Mini Kit (Qiagen, Valencia, CA). Whole transcriptome amplification system (WTA2) (Sigma Aldrich, St. Louis, MO) was used to synthesize μg quantities of amplified cDNA starting with~50 ng of total RNA. Quantitative PCR was measured in SYBR Green/ROX qPCR master mix (Fermentas, Glen Burnie, MD) with primers (IDT, Coralville, IA) shown in S1 Table. qPCR was performed using 7300 Real Time PCR (Applied Biosystems, Foster City, CA). The relative expression level of genes was shown in fold change as normalized to the house keeping gene, eef1a1l1 by using the Ct method. and was shown as means ± SEM (standard error of the mean). The number of BECs was also counted using the ImageJ software as previously described [38]. Unpaired two-tailed Student's t-test was used for statistical analysis; p<0.05 was considered statistically significant.

GWAS identifies ARF6 as a BA susceptibility locus
All DNA samples had excellent data quality with call rates >98%. Of 50 CHP and 30 CAG BA cases and 2818 controls, 39 CHP and 24 CAG cases, and 1907 controls clustered together on principal component analysis of >550000 SNPs, and demonstrated a very low genomic inflation factor of 1.00627 ( Fig 1A) [39]. In comparison of CHP cases with controls, the 1,000 most significant SNPs were ranked further by their proximity to other significant SNPs in the top 1000 in 10 kb windows, to boost confidence in the association (S2 Table). rs3126184 and rs10140366, which ranked 7th and 8th overall in significance emerged as the top-ranked pair demonstrating significantly higher MAFs in CHP cases, in CAG cases, and 63 combined cases from both centers, compared with controls (Table 1). Fourteen additional BA cases genotyped at rs3126184 demonstrated MAF 0.286, and enhanced the significance of the association further, when added to 63 cases above (p = 4.19 x10 -8 , OR = 2.66). rs3126184 and rs10140366 map to 14q22.1 at positions 49442285-49445256 in linkage disequilibrium with each other and the 8.763 kb upstream gene, ARF6 ( Fig 1B). ARF6 regulates embryonic liver development in mice [40]. ARF6 expression is negatively correlated with the minor alleles of rs1040366 (r = -0.1346, p = 0.023) and rs3126184 (r = -0.1128, p = 0.061) in the combined HapMap populations from SNPexp v1.2 [41]. Together, these findings suggest that ARF6 protein expression may be impaired in BA liver tissue.

GWAS pathway analyses further support the importance of ARF6 and EGFR signaling
Among the 14 SNPs that fall into the coding region of genes, two mapped to the genes, CREB3L4 and CRTC2. These genes are related to cAMP response element-binding protein (CREB) pathway, which can be activated by EGFR signaling through the ERK/MAPK pathway [42,43,44,45]. We also performed network and enrichment analyses using Cytoscape and Ingenuity IPA SNP-to-gene mapping in 20kb+/-windows revealed 299 unique genes associated with 419 of the 1000 to-ranked SNPs (S2 Table). Ingenuity's upstream regulator analysis on the 2506 genes (S3 Table) from the first-neighbor network created in Cytoscape revealed that out of 632 potential upstream regulators, the 35 th ranked regulator, sorted by enrichment p-value, was EGF (p = 1.26 x10 -7 ). Other significant EGF-related upstream regulators were TNF (p = 3.14 x10 -25 ) and ERK1/2 (p = 3x10 -14 ). The significant canonical pathways were ERK/ MAPK (p = 5.74 x10 -37 ) and cAMP-mediated signaling (p = 1.5 x10 -35 ) pathways while enriched functional categories were cellular proliferation (p = 1.85 x10 -103 ) and cellular development (p = 1.63 x10 -46 ).

Poor ARF6 staining in human BA liver explants with intrahepatic bile duct paucity
Explants from 29 of 39 CHP cases were immunostained for ARF6 with positive and negative controls. Hepatocytes and bile ducts showed fine granular immunostaining in normal liver tissue (Fig 1Ci). Poor ARF6 staining was seen in two explants with bile duct paucity with minimal fibrosis (Fig 1Cii). In one of these explants, paucity was confined to the left lobe. The right lobe of this and the remaining 27 explants with diffuse ductal hyperplasia, cholestasis, and fibrosis, demonstrated ARF6 staining which was similar to normal liver tissue (Fig 1Ciii). Intense staining was observed in hepatocellular carcinoma (positive control, Fig 1Civ). Therefore, ARF6 protein expression was impaired in the presence of bile duct paucity.
The zebrafish orthologues of human ARF6, arf6a, and arf6b are expressed in the developing liver and endoderm-derived organs Two arf6 genes, arf6a and arf6b, with identical amino acid sequences, are located on chromosomes 20 and 17, respectively, in zebrafish; their amino acid sequences are 99 percent identical to human ARF6. Using an in situ probe for the arf6a coding region whose nucleotide sequence is 81% identical to that of arf6b (S1 Fig, black box), we found that arf6a/b expression appeared ubiquitous at 30 hours post-fertilization (hpf), when the liver-forming region consists of hepatoblasts, but is restricted to the liver at 42 hpf, when hepatoblasts differentiate to biliary epithelial cells (BECs) [46] (Fig 2A, arrows). As embryos further develop, arf6a/b expression was clearly observed in the liver and other endoderm-derived organs including the pancreas and the intestinal bulb (Fig 2A). We used arf6a and arf6b in situ probes targeting non-identical 3'UTR regions of these genes to determine their relative liver expression (S1 Fig, green and red boxes, respectively). Both genes were similarly expressed in the liver and other endoderm-  derived organs at 72 hpf (Fig 2B). At 96 hpf, arf6b expression persisted whereas arf6a expression declined in the liver (Fig 2B).

arf6 knockdown results in biliary defects
To investigate the role of both genes in liver development, we designed three translation-blocking antisense morpholino oligonucleotides (MO). arf6a-UTR and arf6b-UTR MOs targeted ARF6, Biliary Dysgenesis and Biliary Atresia non-identical 5'UTR sequences of corresponding genes; arf6-ATG MO targeted the arf6a ATG start codon region which demonstrated high sequence homology with the start codon for arf6b (S1 Fig, underlines). Their efficacy was evaluated by examining green fluorescent protein (GFP) expression from the CMV:GFP constructs containing each MO-target sequence in front of the GFP start codon (S2 Fig). The arf6-ATG MO blocked GFP expression from the CMV: GFP construct containing the corresponding arf6a or arf6b sequences (S2B Fig) because of high sequence homology in the start codon regions of these two genes, and was therefore used to knock down both genes. The arf6-ATG MO was further validated by whole-mount immunostaining with anti-Arf6 antibody. Arf6 expression was greatly reduced in the MO-injected embryos and increased in arf6a mRNA-injected embryos (S3 Fig). The Tg(Tp1:GFP) um14 line, which expresses GFP in BEC [46,38], and the Tg(fabp10a:dsRed) gz15 line, which expresses dsRed in hepatocytes [47], were used to evaluate intrahepatic biliary structures and liver size, respectively. Liver size was smaller and importantly intrahepatic biliary ducts were relatively sparse in arf6-ATG MO-injected larvae, compared with controls ( Fig 2C). These phenotypes were rescued with arf6a mRNA in a significant number of larvae (S4A Fig). We next examined whether hepatocyte secretion into bile ducts was defective in arf6-ATG MO-injected larvae using the fluorescently labeled fatty acid reporter PED-6, which accumulates in the gallbladder after biliary secretion [37]. PED-6 accumulation in the gallbladder was greatly reduced in the MO-injected larvae compared to controls (Fig 2D). This defect was rescued with arf6a mRNA in a significant number of larvae (S4B Fig). Single injection of arf6a-or arf6b-UTR MO produced biliary defects in a smaller number of larvae than co-injections of these two UTR MOs, or injection with arf6-ATG MO (S5 Fig), indicating that both arf6a and arf6b regulate biliary development.
Detailed characterization of biliary developmental defects was performed at multiple stages from 40 hpf to 5 days post-fertilization (dpf) using whole-mount immunostaining and confocal microscopy. Compared with controls, arf6-ATG MO-injected embryos/larvae demonstrated 1) smaller liver size, as assessed by the expression of the hepatoblast marker Prox1 [48] or the hepatocyte marker fabp10a:dsRed (Fig 3A), 2) reduced number of BECs, as assessed by Tp1: GFP expression, with reduced dispersion and enhanced clustering (Fig 3A), 3) reduced filopodial protrusion of BECs at 72 and 96, but not 62, hpf in fixed larvae ( Fig 3B) and from 74 to 84 hpf in live larvae (S6 Fig), 4) reduced length of interconnecting bile preductules at 5 dpf, as represented by the distance between two nuclei of BECs (Fig 3C). In particular, although liver size in the MO-injected embryos at 62 hpf was similar to that in controls at 48 hpf, their BEC distribution was very different from each other, suggesting that the biliary defects observed in the MO-injected embryos were not simply due to developmental delay. We further confirmed the clustering phenotype, using the Tg(Tp1:H2B-mCherry) s939 line that expresses mCherry in BEC nuclei [38]. The quantification of the percentage of single BECs versus cells in cluster of two, three, or four and more cells revealed a significant decrease in the percentage of BECs present as single cells and doublets in the MO-injected larvae compared to controls, concomitant with a significant increase in the percentage of cells in clusters of four and more cells (Fig 3D). This clustering phenotype was not due to increased BEC proliferation. In a proliferation assay, the rate of BEC proliferation was not significantly different between controls and the MO-injected larvae at 75 hpf (S7C Fig), although the total number of BECs in each liver was smaller in the MO-injected larvae than in controls (S7B Fig). Since bile first secretes into bile canaliculi that connect bile preductules in zebrafish and bile ductules in mammals, we also examined the expression patterns of Abcb11, a bile salt export pump present in the bile canaliculi of hepatocytes [49]. The canaliculi were much shorter in the MO-injected larvae than in controls ( Fig  3E). Since arf6a and arf6b are also expressed in other endoderm-derived organs (Fig 2B), we investigated whether other endoderm-derived organs are defective in arf6-ATG MO-injected embryos using the Tg(sox17:GFP) s870 line that expresses GFP in all endodermal cells [32]. The pancreas and the intestinal bulb were smaller in the MO-injected embryos than in controls (S8A Fig, arrowheads and brackets, respectively). In addition, the budding of the swim bladder, the zebrafish equivalent of the mammalian lung, was not evident in the MO-injected embryos at 48 hpf, whereas it was evident in controls ( S8A Fig, open arrowhead). In contrast to these defects, the gallbladder, EHBD, and the extrapancreatic duct appeared normal in the MOinjected larvae at 75 hpf (S8B Fig), as assessed by the 2F11 antibody, which targets the hepatopancreatic ductal system [35].
We next investigated whether EGFR signaling and Arf6 acted in the same pathway to regulate intrahepatic biliary morphogenesis. At four-fold lower doses (0.5 ng and 1 μM, respectively), arf6-ATG MO or AG1478 alone produced minimal or no defects in intrahepatic biliary structure at 75 hpf or in PED-6 accumulation in the gallbladder at 5 dpf. When these two reagents were combined, biliary structure was greatly affected (Fig 5A and 5B) and PED-6 accumulation greatly reduced (Fig 5C), similar to embryos injected with 2 ng of arf6-ATG MO (Fig 2C and 2D). These data suggest that EGFR and Arf6 likely regulate intrahepatic biliary morphogenesis via the same pathway.
EGFR-GEP100-ARF6 pathway genes are upregulated in arf6 MOinjected larvae We next examined, by quantitative PCR, the expression levels of EGFR-GEP100-ARF6 pathway genes. ARF6 is necessary for membrane pseudopod formation via ASAP1 [52] and RAC1 and also activates ERK/MAPK signaling via RAC1 [53,54], We also examined the expression of hedgehog signaling genes which were upregulated in zebrafish embryos after knockdown of gpc1 in a previous study [13].
Dissected liver tissue from 4.5-dpf zebrafish larvae, injected with arf6-ATG MO in three batches of~100 larvae each, showed that egfra, gep100, arf6, asap1 and rac1 were upregulated in all replicates in arf6 morphants compared with uninjected embryos (Fig 6). Among the hedgehog signaling genes, only ptch1, but not gli1 or gli2a demonstrated consistent upregulation in all three replicates (S4 Table). The transcription factors, gli1, and gli2a mediate the effects of the hedgehog receptor ptch1 on downstream genes in the canonical hehdgehog signaling pathway. These observations suggested that arf6 knockdown affects EGFR, but not canonical hedgehog signaling in the liver.

Discussion
Our study identifies the chromosome 14q21.3 locus, defined by the highly associated SNPs rs3126184 and rs10140366, as a susceptibility locus for BA. This 3' flanking region includes the upstream ARF6 gene, because minor alleles at both SNP loci are associated with decreased ARF6 expression in the combined HapMap populations (SNPexp v1.2), and rs10140366 lies in an enhancer region enriched for the histone marks H3K4me1 and H3K27Ac (Fig 1B). Further, Arf6 -/knockout mice show impaired liver development [40]. Although distinct from GPC1 and ADD3, two genes associated with BA in recent GWAS, the ARF6 gene has similar functional and developmental consequences [13,16]. Knockdown of both gpc1 and arf6 in zebrafish impairs biliary network formation. This is expected because GPC1 and ARF6 mediate fibroblast and epidermal growth factor signaling, respectively, which are necessary for the growth and development of organs by epithelial branching morphogenesis [55,56]. ADD3 may also have a developmental role because it is expressed to a greater extent in fetal liver than in adult liver [57]. ARF6 and ADD3 both regulate actin cytoskeletal remodeling, and affect cell mobility and cell-cell contact, which facilitate organogenesis [52,58]. Among the top-ranked 1000 SNPs in our study, rs17127145 closest to the ADD3 gene ranked 676 th ; none was in proximity to the glypican-1 gene. Therefore, the association between bile duct paucity and poor ARF6 immunostaining in explants with BA led us to further ask whether arf6 knockdown in zebrafish could affect early biliary morphogenesis and explain the clinical manifestations of BA. Poor intrahepatic biliary network formation and poor bile excretion manifested as absence of PED-6 accumulation in the gallbladder are dominant effects of arf6 knockdown. Clinical correlates include failure to excrete radionuclide into the extrahepatic bile duct on hepatobiliary scanning, failure of surgical drainage, and bile duct paucity in some BA patients [59,60]. In arf6-ATG MO-injected zebrafish embryos, poor bile excretion also results from reduced biliary canalicular length and enhanced clustering of BEC, presumably due to reduced BEC mobility during morphogenesis. Altered epithelial cell polarity is a known effect of ARF6 deficiency in MDK canine kidney cell line [61]. Clinical correlates include canalicular dilatation with cystic bile lakes characteristic of the "ductal plate malformation" in up to a fourth of BA patients [62]. Because MO is diluted out by cell division, arf6 MO effects are short-lived. arf6 mutant lines are needed to investigate long-term consequences. Other pathways may also function at subsequent developmental stages and explain why EHBD loss and atretic gallbladders characteristic of human BA were absent in arf6 MO-injected zebrafish.
ARF6 is a critical member of the EGFR pathway, which facilitates organ development, growth and regeneration by regulating epithelial branching morphogenesis [56]. ARF6 is activated by the binding of EGFR to its activator, GEP100 and has several downstream effects. Activation of the downstream effector ASAP1 is necessary for a normal actin cytoskeleton and cell membrane function [52]. An upregulated EGFR-GEP100-ARF6 pathway facilitates invadopodia formation in invasive breast cancer. ARF6 is also required for RAC1-mediated epithelial cell polarity [61] and membrane pseudopod formation [53]. RAC1 activation also regulates ERK signaling [54]. Sequential activation of ERK/MAPK and CREB signaling pathways is required for normal cellular development and proliferation, and is another downstream effect of EGFR-GEP100-ARF6 signaling [43,44,45,52]. Systems biology analysis of GWAS results supports the role of this molecular pathway in BA by showing enrichment of EGF regulatory genes, the canonical ERK/MAPK and CREB signaling pathways, and the functional categories of cellular development and proliferation (Fig 7).
Therefore, we hypothesized further and found that EGFR signaling blockade with the inhibitor AG1478 reproduced all biliary defects associated with arf6-ATG MO-induced knockdown in zebrafish. Further, reduced doses of AG1478 and arf6-ATG MO also reproduced these defects when used in combination, but not when used alone. These observations suggest that EGFR-ARF6 signaling regulates early branching morphogenesis of the biliary tree, and that functionally impaired pathway members could also contribute to poor bile drainage likely from abnormal intrahepatic biliary network in children with BA. Supportive evidence consists of poor ARF6 immunostaining in those liver explants, which manifested bile duct paucity without fibrosis, but not in the remainder, in which advanced cirrhosis with marked cholestasis precluded an evaluation of intrahepatic bile ducts prior to the development of chronic obstructive changes ( Fig 1C).
The complex molecular effects of arf6 knockdown in zebrafish also provide potential explanations for the fibrosis that is seen in BA. arf6 knockdown in zebrafish was associated with intrahepatic upregulation of egfra, gep100, arf6, and genes for downstream effectors of Arf6 such as asap1 and rac1 (Fig 6). Among the canonical hedgehog signaling pathway genes, which have been implicated previously in biliary dysgenesis only ptch1, which induces several hedgehog target genes via effects on the gli family of receptors, and not the gli1 or gli2 genes, was consistently upregulated in arf6 morphants (S4 Table) (13). Activation of Erk by sonic hedgehog can occur independent of canonical hedgehog signaling via PTCH [63] further confirming that ARF6 dysregulation may have several downstream effects on cell membrane function, cell proliferation and cell development via effects on downstream effectors, ASAP1, RAC1 and PTCH1. In experimental models, upregulated EGFR pathway members lead to fibrosis and cirrhosis, a common finding at liver transplantation after failed portoenterostomy for BA [64].
These observations from an established zebrafish model of bile duct development provide additional supportive evidence for the biological relevance of ARF6 in biliary atresia in this first report. Obvious future directions also include replication of these experimental findings. This task requires that the effects of an induced perturbation be observed over several days in ARF6, Biliary Dysgenesis and Biliary Atresia invitro mammalian cell lines, because complete disruption of the ARF6 gene in Arf6 -/-mice leads to almost complete lethality [40]. Because previous studies conducted in patients with BA indicate a high likelihood of multiple susceptibility loci and the environment in disease pathogenesis, future models, which cannot demonstrate such additive effects are likely to maintain a supportive role placing the larger burden of proof back on the human evidence from diseased subjects.
It is generally assumed that the intrahepatic biliary tree is also abnormal in extrahepatic atresia [65,59,60]. However, intrahepatic ducts cannot be examined before the onset of jaundice and clinical evidence of extrahepatic obstruction in human BA. Anecdotally, serial biopsies from four children with confirmed extrahepatic biliary atresia have demonstrated that intrahepatic bile duct paucity without fibrosis preceded obstructive ductal proliferation and cirrhosis at portoenterostomy or liver transplantation [66]. Features indicative of intrahepatic duct involvement are predominantly clinical; disease progression despite portoenterostomy, an attenuated intrahepatic picture on cholangiography; and histopathologic-lymphocytes around and within intrahepatic BECs [67,68]. In the Rhesus rotavirus mouse model of BA, intrahepatic portal areas may contain lymphocytes though the relevance of the model to human BA is unclear [11,69]. Also not evident is whether intrahepatic changes in human BA are primary, or secondary to extrahepatic obstruction. Landing, in 1974, proposed a unitarian hypothesis of infantile obstructive cholangiopathies in which the entire biliary tree was vulnerable but extrahepatic or intrahepatic disease depended on the developmental timing and nature of the insult [70]. The concept lapsed for lack of proof until now. The zebrafish model from two independent studies, including ours, suggests again that the entire biliary tree is genetically susceptible and that in this instance, the outcome of ARF6 dysregulation involves the intrahepatic tree.
In summary, GWAS studies and analysis of SNP data for functions and pathways postulates the role of ARF6, involved in the EGFR pathway, in defective bile duct formation. We validate this hypothesis in a zebrafish model, which unambiguously shows that arf6 knockdown impacted the EGFR pathway in the liver and lead to defective bile duct formation. Wild-type embryos were injected at the one-cell stage with 2 ng of arf6-ATG MO or 120 pg of arf6a mRNA, harvested at 7 hpf, and processed for whole-mount immunostaining with anti-Arf6 (green) and anti-β-catenin (red) antibodies. DNA was also stained with Hoechst 33342 (gray). β-catenin expression reveals the cell membrane. Confocal images showed that Arf6 expression was greatly increased in arf6a mRNA-injected embryos, whereas it was greatly reduced in arf6-ATG MO-injected embryos, validating the efficacy of the MO. Dorsal views; scale bar, 25 μm. (TIF) S4 Fig. arf6a mRNA injection partially rescues biliary defects in arf6-ATG MO-injected larvae. (A) The Tg(Tp1:GFP), Tg(fabp10a:dsRed), and Tg(ins:dsRed) lines were used to reveal the intrahepatic biliary structure, the liver, and the dorsal pancreas, respectively. Epifluorescence images showing the expression of these transgenes revealed that a defect in the intrahepatic biliary structure in arf6-ATG MO-injected larvae was partially rescued by arf6a mRNA injection. Based on the severity of the biliary defect, larvae were divided into three groups: normal, intermediate, and severe. Confocal images showing the endoderm and endoderm-derived organs at 48 hpf. The Tg (sox17:GFP) line was used to reveal the endoderm and endoderm-derived organs. Arrows, arrowheads, and open arrowhead point to the liver, the pancreas and the swim bladder, respectively; brackets mark the intestinal bulb. (B) Confocal images showing the hepatopancreatic ductal system at 75 hpf. Tg(Tp1:GFP);Tg(fabp10a:dsRed) larvae were processed for wholemount immunostaining with 2F11 (gray), GFP (green), and dsRed (red) antibodies. In contrast to the intrahepatic biliary defect, the hepatopancreatic ductal system, revealed by 2F11 antibody, appeared to be normal in arf6-ATG MO-injected larvae. Arrows, arrowheads, open arrowheads, and open arrows point to the gallbladder (gb), the extrahepatic duct (ehd), the common bile duct (cbd), and the extrapancreatic duct (epd), respectively. The hepatopancreatic ductal system is schematically illustrated. Ventral views, anterior up. Scale bars, 50 μm. (TIF) S1 Table. qPCR primer sequences for gene expression studies in liver tissue from zebrafish. (DOCX) S2 Table. Four hundred nineteen SNPs mapped to unique 299 genes within 20kb+/-windows among the top 1000 significant SNPs in CHP cohort with BA in the genome wide association study. (XLS) S3 Table. Among the 1000 top-ranked significant SNPs, 419 mapped to 299 unique genes as described in the methods. Among these genes, 231 showed interaction with other genes from the large human network consisting mainly of protein-protein interaction. The first neighbor network was created by assimilating all of the first interacting genes with these 231 genes from the large human network. The total of 2506 genes in the new smaller network is listed in the table. More information can be found in the table that includes the average P-values calculated from multiples SNPs and the interacting genes. (XLSX) S4 Table. Quantitative RT-PCR analysis of hedgehog target genes in liver tissue from zebrafish.