NCK-associated protein 1 like (nckap1l) minor splice variant regulates intrahepatic biliary network morphogenesis

Impaired formation of the intrahepatic biliary network leads to cholestatic liver diseases, which are frequently associated with autoimmune disorders. Using a chemical mutagenesis strategy in zebrafish combined with computational network analysis, we screened for novel genes involved in intrahepatic biliary network formation. We positionally cloned a mutation in the nckap1l gene, which encodes a cytoplasmic adaptor protein for the WAVE regulatory complex. The mutation is located in the last exon after the stop codon of the primary splice isoform, only disrupting a previously unannotated minor splice isoform, which indicates that the minor splice isoform is responsible for the intrahepatic biliary network phenotype. CRISPR/Cas9-mediated nckap1l deletion, which disrupts both the primary and minor isoforms, showed the same defects. In the liver of nckap1l mutant larvae, WAVE regulatory complex component proteins are degraded specifically in biliary epithelial cells, which line the intrahepatic biliary network, thus disrupting the actin organization of these cells. We further show that nckap1l genetically interacts with the Cdk5 pathway in biliary epithelial cells. These data together indicate that although nckap1l was previously considered to be a hematopoietic cell lineage-specific protein, its minor splice isoform acts in biliary epithelial cells to regulate intrahepatic biliary network formation.


Introduction
The biliary system is responsible for transporting bile from the liver to the digestive tract. The biliary system within the liver, the intrahepatic biliary network, is a highly branched threedimensional network that is found throughout the liver. Biliary epithelial cells, also called cholangiocytes, are the innermost epithelial cells forming the conduit of this network. Threedimensional branching anomalies in the intrahepatic biliary network are implicated in many liver diseases, including biliary atresia [1,2]. Biliary atresia is the most common neonatal cholestatic disorder, occurring in approximately 1 of 18,000 live births [3], but the molecular etiology and pathogenesis of this disease are poorly understood. Biliary atresia is not generally thought of as an inherited disease; however, there are some lines of evidence to support the hypothesis that genetic factors contribute to susceptibility to this disease [4][5][6]. However, the genes and signaling pathways underlying genetic susceptibility to biliary atresia affecting intrahepatic biliary network branching morphogenesis have not been fully investigated. Dysregulation of the innate immune system has been proposed to be involved in the pathogenesis of biliary atresia [7]. However, the question of whether genes important for the innate immune system also affect biliary system formation has just begun to be investigated.
We have previously used a pharmacogenetics approach to show that the Cdk5-Pak1--Limk1-Cofillin kinase cascade regulates branching morphogenesis of the intrahepatic biliary network and actin dynamics in biliary epithelial cells [8]. However, the extent to which this Cdk5-mediated signaling pathway is linear or branched to regulate actin dynamics is not known.
Alternative splicing is a general regulatory mechanism that produces more than one mRNA isoform from a single gene, allowing the generation of different protein isoforms with diverse functions or localizations from a single gene [19]. Approximately 90-95% of human genes are known to undergo alternative splicing [20]; however, further work is required to uncover the physiological functions of each of those splice isoforms. Some cardiac and neuronal developmental human diseases are associated with errors in alternative splicing mechanisms [19]. However, it is not known whether any alternative splicing is involved in biliary system formation.
In this study, through N-ethylnitrosourea (ENU)-mediated forward genetic screening in zebrafish combined with computational intrahepatic biliary network structural analysis, we have identified a new mutation that genetically interacts with the Cdk5-mediated pathway during branching morphogenesis. We identified that the mutation disrupts a previously unannotated minor splice isoform of the nckap1l gene, and propose that this minor isoform of Nckap1l functions downstream of Cdk5 and Rac1 to regulate branching morphogenesis of the intrahepatic biliary network.

A forward genetic screen identified the lri35 allele showing a specific phenotype in the intrahepatic biliary network
In the course of ENU-based mutagenesis utilizing Tg(Tp1-MmHbb:EGFP) um14 expression [21,22] in the intrahepatic biliary network and its computational network structure analysis [8], we identified a recessive mutant, lri35, which shows a phenotype similar to that of Cdk5-suppressed larvae [8] (Materials and Methods). The lri35 mutation is viable, and the physical appearance of the majority of lri35 mutant larvae is almost indistinguishable from that of wild-type siblings (Fig 1A and 1B) at 5 days post-fertilization (dpf), while approximately 23% of lri35 mutant larvae (11/47 mutant larvae from three independent crosses) show a delay in swim bladder inflation at this stage. At this stage, the body length of lri35 mutant larvae (average 3.87 mm, s.d. = 0.20, n = 11) is not significantly different from that in wild-type siblings (average 3.96 mm, s.d. = 0.12, n = 14, p = 0.17) at 5 dpf [23], and the size of the liver (Materials and Methods) remains unchanged. However, at this stage, the intrahepatic biliary network in lri35 mutant larvae appears to show fewer branches (Fig 1C and 1D). To quantify the difference in the branching pattern between wild-type and lri35 mutant larvae, we utilized the computational network structure analysis and measured the structural properties of the intrahepatic biliary network (Fig 1E and 1F). We found that network volume ( Fig 1G) and length ( Fig 1H) in lri35 mutant larvae are reduced while network average thickness is increased (Fig 1I), indicating that the intrahepatic biliary network becomes thicker and shorter in lri35 mutant larvae. In lri35 mutant larvae, the number of nodes (Fig 1J), node-node connections (Fig 1K), and unconnected branches (node-endpoint connections) ( Fig 1L) are all reduced while the ratio of connected and unconnected branches is increased (Fig 1M), suggesting that new branches fail to form connections properly in lri35 mutant larvae. The number of Tg(Tp1-MmHbb:EGFP) um14 expressing cells in the liver is reduced in lri35 mutant larvae at 5 dpf (Fig 1O), suggesting that the number of BECs is reduced. Bile canaliculi are the apical membranes of hepatocytes that connect to the intrahepatic biliary network. In lri35 mutant larvae, bile canaliculi, as marked by Abcb11 expression [8], remain relatively unaffected ( Fig 1P and 1Q), and the average canalicular length is not changed (Fig 1R). The density of canalicular connection to the intrahepatic biliary network is increased in lri35 mutant larvae (Fig 1S), possibly due to the reduction of the total network length of the intrahepatic biliary network. Overall, these data together indicate that the lri35 mutation impacts branching morphogenesis of the intrahepatic biliary network in a relatively specific manner.

The lri35 mutation disrupts the minor splice isoform of nckap1l
Until recently, mapping of a causative mutation of ENU mutagenesis-derived mutants was an extremely labor-intensive process. However, the development of next-generation sequencing- Lateral views. At 5 dpf, there is no significant difference in physical appearance in nckap1l lri35 mutant larvae, although their swim bladder inflation is occasionally delayed. (C and D) Projected confocal images of Tg(Tp1-MmHbb: EGFP) um14 expression in WT (C) and nckap1l lri35 mutant (D) larvae at 5 dpf. GFP expression in the intrahepatic biliary network is shown in pseudocolored yellow. Ventral views, anterior to the top. (E and F) Skeletal representation of the intrahepatic biliary network in WT (E) and nckap1l lri35 mutant (F) larvae computed based on Tg(Tp1-MmHbb: EGFP) um14 expression at 5 dpf. The complex three-dimensional network is represented by a combination of four segments: end points (colored green), nodes (colored white), node-node connections (colored red), and node-end point connections (colored yellow). (G-M) Computational skeletal analysis-based measurements of the intrahepatic biliary network structures of WT and nckap1l lri35 mutant larvae at 5 dpf. (G) The total network volume of the intrahepatic based approaches [24][25][26] has made this process significantly easier. We applied the MMAPPR algorithm [24] to RNA-seq data obtained from lri35 mutant larvae at 5 dpf (Materials and Methods) to map the location of the causative mutation. The lri35 mutation was mapped to chromosome 11 at a region closer to the distal tip (Fig 2A and 2B). Based on the RNA-seq data, we first examined whether there were any SNPs inducing a premature stop within the critical region, but we did not detect any of these mutations. We next screened for differential gene expression levels and differential exon usage of genes within the critical region by using the Cufflinks algorithm [27]. We found that the most down-regulated gene within the critical region (yet bigger than a half-fold change) in lri35 mutant larvae was nckap1l. Based on these results, we hypothesized that the lri35 mutation might disrupt the nckap1l gene. To test this hypothesis, we first injected antisense morpholino (MO) against nckap1l into wild-type Tg (Tp1-MmHbb:EGFP) um14 eggs (Materials and Methods). We found that at 5 dpf, larvae injected with nckap1l MO did not show any changes in physical appearance, but did show a specific phenotype in the intrahepatic biliary network (S1 Fig) consistent with the hypothesis that nckap1l is mutated in lri35 mutant larvae. We next induced an indel mutation to the nckap1l gene in wild-type using CRISPR/Cas9-based genome editing technology [28,29]. We injected the assembled CRISPR/Cas9 complex targeting nckap1 into fertilized zebrafish eggs (Materials and Methods) and subsequently established a deletion allele, nckap1l lri90 , that deletes 7 bp from the first exon and induces a premature stop at the position of 3  The nckap1l lri90 mutation is viable, and homozygous nckap1l lri90 adult fish show no overt difference in physical appearance. However, the Tg(Tp1-MmHbb:EGFP) um14 positive network of the homozygous nckap1l lri90 adult liver appears to be thinner and less dense (S2U Fig) than that of wild-type fish, suggesting that the mutation continues to affect the intrahepatic biliary network in the adult stage. To test the hypothesis that the lri35 mutation disrupts the nckap1l gene, we attempted a genetic complementation approach by crossing heterozygous lri35 and homozygous nckap1l lri90 mutant fish. We found that approximately 50% (20/43) of the larvae from this cross show a phenotype indistinguishable from either that of lri35 or nckap1l lri90 mutant larvae (S3 Fig), indicating that the lri35 allele failed to complement the nckap1l lri90 allele and that these two mutations disrupt the same gene. However, to our surprise, we did not find any mutation when we sequenced the nckap1l cDNA isolated from lri35 mutant larvae at 5 dpf  Whole genome-wide Loess fit curve for SNP allele frequency Euclidean distance computed by the MMAPPR algorithm. Chromosome 11 showed the highest score. (B) Loess fit curve for SNP allele frequency Euclidean distance within chromosome 11. The critical region was mapped to the distal tip of chromosome 11. (C) The nckap1l gene has two splice isoforms: the previously annotated major splice isoform (α isoform) and an unknown minor isoform (ß isoform). The lri35 mutation, which is a one-nucleotide insertion, is located in the last exon and induces a frameshift only in the ß isoform of the nckap1l gene. These data indicate that the minor isoform of the nckap1l gene is responsible for the phenotypes in lri35 mutant larvae. (D) Schematic of Nckap1l α and ß proteins. The entire Nckap1l α protein is recognized as the Nckap1l domain (purple box). Due to the alternative splicing of the nckap1l gene, Nckap1l ß shifted to the ß specific sequence (orange box) after the position 579. The lri90 mutation affects both Nckap1l α and ß, whereas the lri35 mutation affects Nckap1l ß only. (E) The lri35 mutation is an insertion mutation that inserts an additional adenine nucleotide into the nckap1l gene. (F) The lri35 mutation induces a frameshift and changes the last 13 amino acids of the ß isoform of the Nckap1l protein.
( Fig 2C). We also sequenced most of the annotated exon/intron boundaries of the nckap1l gene in lri35 mutant larvae but did not find any mutations. By chance during these intensive sequence attempts, we found a previously unannotated splice isoform of nckap1l, which we refer to as nckap1l ß. This alternative splice variant skips all exons after exon 16 and uses an alternative stop codon in the last exon ( Fig 2C). Thus, the encoded Nckap1l ß protein is smaller than Nckap1l (Fig 2D). We isolated the nckap1l ß cDNA from lri35 mutant larvae and found that one adenine nucleotide is inserted into the ß-specific coding region (Figs 2C, 2E and S4). This insertion changes the last 13 amino acids of the Nckap1l ß protein (Fig 2F). We measured relative nckap1l α and ß gene expression in lri35 mutant larvae at 5 dpf by quantitative RT-PCR and found that nckap1l ß expression is slightly reduced in lri35 mutant larvae whereas nckap1l α expression remains unchanged (S5A and S5B Fig). Since the entire nckap1l ß gene sequence is part of the ORF and UTR of the nckap1l α gene, we were not able to design the nckap1l ß-specific RNA probe for in situ hybridization. Instead, we compared nckap1l αspecific probe to nckap1l α and ß shared probe, and found that the nckap1l α and ß shared probe stained the liver more strongly than the nckap1l α-specific probe, suggesting that nckap1l ß might be expressed more than nckap1l α in the liver at 5 dpf (S5C and S5D Fig).
Based on these data, we propose that the lri35 mutation is an adenine nucleotide insertion that specifically disrupts the ß isoform of the nckap1l gene, indicating that this minor splice isoform of nckap1l is important for intrahepatic biliary network branching morphogenesis.

Nckap1l is expressed in hepatic biliary epithelial cells
In mice, Nckap1l is predominantly expressed in immune cells [16]; however, its expression within biliary epithelial cells has not been studied. We generated an antibody against zebrafish Nckap1l (Materials and Methods) and examined Nckap1l expression in the zebrafish larval liver at 5 dpf. We found that Nckap1l is predominantly expressed in vascular endothelial cells (Figs 3A and S6). We also observed small puncta of Nckap1l expression in Tg(Tp1-MmHbb: EGFP) um14 -positive biliary epithelial cells ( Fig 3B). Expression in both endothelial and biliary epithelial cells is missing in nckap1l lri35 mutant larvae at 5 dpf (Figs 3C and S7), suggesting that the Nckap1l protein is degraded in the mutant larvae.

WAVE regulatory complex component proteins are degraded in biliary epithelial cells in the liver of nckap1l lri35 mutant larvae
In other systems, depletion of Nckap1l leads to the destruction of other WAVE regulatory complex (WRC) component proteins (Fig 3D), including WAVE and Abi1, without influencing their level of gene expression [16,18,[31][32][33]. We found that WAVE1 is expressed predominantly in biliary epithelial cells in the liver of wild-type larvae at 5 dpf (Fig 3E), and this WAVE1 expression is decreased in nckap1l lri35 mutant larvae ( Fig 3F). Moreover, in the wildtype liver at 5 dpf, consistent with a previous report that Abi1 undergoes nucleocytoplasmic shuttling [34], Abi1 is localized to the nucleus of some hepatocytes and biliary epithelial cells ( Fig 3G). In nckap1l lri35 mutant larvae, Abi1 expression in biliary epithelial cells is specifically missing, while its expression in hepatocytes remains unchanged ( Fig 3H). Similarly, we found that HSPC300 is expressed in the liver, including in biliary epithelial cells, of wild-type larvae at 5 dpf (Fig 3I), and this HSPC300 expression is decreased in nckap1l lri35 mutant larvae ( Fig  3J). Three-dimensional projected confocal images of the liver in wild-type larvae indicated that Nckap1l and HSPC300 colocalize in biliary epithelial cells near the projection tips of biliary epithelial cells ( Fig 3K) at 5 dpf, suggesting that the WRC forms at this site. These data together indicate that WRC forms in biliary epithelia cells in wild-type larvae, and WRC component proteins are degraded in the liver of nckap1l lri35 mutant larvae. . This complex acts downstream of Cdk5 and activated Rac1 to stimulate actin remodeling through the actin regulatory complex. The WAVE regulatory complex is a pentameric heterocomplex that consists of WAVE (1, 2 or 3), Abi (1 or 2), Sra1, Nckap1l (or Nap1), and HSPC300. (E and F) Z-plane confocal images of the liver visualized for WAVE1 expression (Red) in WT (E) and nckap1l lri35 mutant (F) larvae at 5 dpf. Overlay images with biliary epithelial cell marker Tg (Tp1-MmHbb:EGFP) um14 are shown separately in (E' and F'). WAVE1 is expressed predominantly in biliary epithelial cells in WT larvae; however, in nckap1l lri35 mutant larvae, WAVE1 expression disappears from biliary epithelial cells, suggesting that WAVE1 undergoes degradation in biliary epithelial cells. (G and H) Z-plane confocal images of the liver visualized for Abi1 expression in WT (G) and nckap1l lri35 mutant (H) larvae at 5 dpf. Overlay images with biliary epithelial cell marker Tg(Tp1-MmHbb:EGFP) um14 are shown separately in (G' and H'). Abi1 localizes to the nuclei of hepatocytes and biliary epithelial cells (white arrowheads) in the wild-type liver. However, in the liver of nckap1l lri35 mutant larvae, Abi1 staining in the nuclei of biliary epithelial cells is lost (yellow arrows), while Abi1 expression Consistent with the observation that WAVE regulatory complex proteins are degraded in biliary epithelial cells, the actin organization of nckap1l lri35 mutant larvae changes and becomes concentrated around the lumen of biliary epithelial cells (Fig 3L and 3M).

Overexpression of nckap1l ß specifically impacts the intrahepatic biliary network
As mutations in the nckap1l ß gene were found to impact intrahepatic biliary network formation, we next examined whether overexpression of nckap1l ß would also influence intrahepatic network formation. We synthesized the nckap1l ß mRNA in vitro (Materials and Methods) and injected it into Tg(Tp1-MmHbb:EGFP) um14 eggs. At 5 dpf, we did not observe any difference in physical appearance in nckap1l ß mRNA-injected larvae (Fig 4A and 4B). We found that all injected larvae appeared to show differences in intrahepatic biliary network branching pattern (Fig 4C-4F). However, likely due to the degradation of injected RNA by 5 dpf, there was high variability of the intrahepatic biliary network structural properties in larvae injected with nckap1l ß mRNA at 5 dpf (Fig 4G-4L), and only the percentage of 5-or-more-way branches among the entire network was increased significantly (Fig 4L), suggesting that the network in nckap1l ß mRNA-injected larvae contains nodes with more branches. These data indicate that overexpression of nckap1l ß can induce a relatively specific phenotype in the intrahepatic biliary network.

Overexpression of nckap1l ß rescued nckap1l lri35 mutant phenotypes
We next injected nckap1l ß mRNA into eggs obtained from an nckap1l lri35 heterozygous fish intercross, and at 5 dpf, we genotyped homozygous nckap1l lri35 larvae injected with the nckap1l ß mRNA and examined intrahepatic biliary network formation. We found that network structural sub-parameters observed in nckap1l lri35 mutant larvae, including the total network length (Fig 4M), the numbers of nodes ( Fig 4N) and connections (Fig 4O), are rescued by injecting nckap1l ß mRNA. These data indicate that supplying nckap1l ß mRNA can rescue, at least in part, intrahepatic biliary phenotypes in nckap1l lri35 mutant larvae, further confirming that the nckap1l ß gene is responsible for this mutation.

nckap1l genetically interacts with Cdk5 and Rac1 during intrahepatic biliary network branching morphogenesis
We identified the nckap1l lri35 mutant as having a similar phenotype to that of Cdk5-suppressed larvae. In the mammalian nervous system, Cdk5 is known to directly phosphorylate WAVE1 remains in hepatocytes, suggesting that Abi1 undergoes degradation specifically in biliary epithelial cells. (I and J) Zplane confocal images of the liver visualized for HSPC300 expression in WT (I) and nckap1l lri35 mutant (J) larvae at 5 dpf. Overlay images with biliary epithelial cell marker Tg(Tp1-MmHbb:EGFP) um14 are shown separately in (I' and J'). Similar to Nckap1l, HSPC300 is predominantly expressed in endothelial cells in the liver, but HSPC300 is also expressed in biliary epithelial cells (White arrowheads in I). HSPC300 expression is missing in the liver of nckap1l lri35 mutant larvae (J). (K) Projected confocal image of co-localizing signal between Nckap1l and Tg(Tp1-MmHbb: EGFP) um14 showing Nckap1l expression (pseudocolored green) in biliary epithelial cells at 5 dpf. Tg(Tp1-MmHbb: EGFP) um14 expression is shown in pseudocolored blue. Nckap1l is localized near the projection tip (white arrowhead), but not exactly at the tip. Colocalizing signal between HSPC300 and Tg(Tp1-MmHbb:EGFP) um14 showing HSPC300 expression (red) in biliary epithelial cells is overlaid in (K'). This indicates that HSPC300 is expressed in biliary epithelial cells. Nckap1l and HSPC300 colocalize in biliary epithelial cells (orange arrowheads) near the projection tips of biliary epithelial cells, suggesting that the WRC forms at this site.  to regulate actin cytoskeletal organization [15]. Thus, we hypothesized that nckap1l and cdk5 might act in the same signaling pathway to regulate intrahepatic biliary network branching morphogenesis. To test this hypothesis, we examined the genetic interaction between nckap1l and cdk5 in biliary epithelial cells. In the liver of Tg(tp1:cdkal1) kl109 larvae, the endogenous inhibitor of Cdk5 is expressed specifically in biliary epithelial cells [35]. Consistent with a previous report that the phenotype of Tg(tp1:cdkal1) kl109 larvae is very mild [35], we found that the branching pattern of the intrahepatic biliary network in Tg(tp1:cdkal1) kl109 larvae is not significantly changed from that of wild-type larvae at 5 dpf (Fig 5A, 5C and 5E). As the nckap1l lri90 mutation is recessive, we did not observe any phenotype in heterozygous nckap1l lri90 mutant animals. However interetingly, when we crossed heterozygous nckap1l lri90 mutant fish to homozygous Tg(tp1:cdkal1) kl109 fish, we found that approximately 50% of their offspring (n = 23/45 tested) showed a significantly more severe phenotype in the intrahepatic biliary network compared to that of Tg(tp1:cdkal1) kl109 larvae at 5 dpf (Fig 5B, 5D and 5F). This result strongly suggests that nckap1l and cdk5 genetically interact in biliary epithelial cells. Subsequent genotyping confirmed that those 23 larvae showing an enhanced phenotype are all heterozygous nckap1l lri90 larvae expressing Tg(tp1:cdkal1) kl109 , confirming that losing one copy of nckap1l in the Tg(tp1:cdkal1) kl109 background synergistically generates the phenotype via the dosage-sensitive genetic interaction. Computational network structure analysis confirmed that network volume, network length, and segment number are significantly reduced in heterozygous nckap1l lri90 ; Tg(tp1:cdkal1) kl109 larvae compared to those in Tg(tp1:cdkal1) kl109 larvae (Fig 5G-5J). These data together indicate that nckap1l genetically interacts with the Cdk5 pathway in biliary epithelial cells to regulate intrahepatic biliary network remodeling. Finally, since the WRC is known to acts downstream of Rac1 [12][13][14]36], we tested whether the nckap1l mutation genetically interacts with the Rac1 pathway. Consistent with a previous report [37], 50 ug/ml Rac1 inhibitor treatment from 3 to 5 dpf did not induce any physical appearance change, but we found that the intrahepatic biliary network branching pattern was severely altered (Fig 6B and 6F). When we lowered the Rac1 inhibitor concentration to 10 ug/ ml, it no longer affected intrahepatic biliary network branching patterns in wild-type larvae (Fig 6D and 6H). However, when we treated larvae obtained from nckap1l lri90 heterozygous outcross to wild-type fish from 3 to 5 dpf with the same low dose 10 ug/ml Rac1 inhibitor, approximately 50% (24/51) larvae showed a significantly more severe phenotype in the intrahepatic biliary network (Fig 6C, 6G, and 6I-6L), and we identified that these affected larvae are all heterozygous nckap1l lri90 larvae. These data together show that losing one copy of nckap1l sensitized the effects of the low dose Rac1 inhibitor, suggesting that nckap1l acts downstream of the Rac1 pathway.  EGFP) um14 expression in nckap1l lri90(+/+) ;Tg(tp1:cdkal1) kl109 (C) and nckap1l lri90(+/-) ;Tg(tp1:cdkal1) kl109 (D) larvae at 5 dpf. GFP expression in the intrahepatic biliary network is shown in pseudocolored yellow. nckap1l lri90(+/-) ;Tg(tp1: cdkal1) kl109 larvae show more severe intrahepatic biliary network phenotypes than nckap1l lri90(+/-) ;Tg(tp1:cdkal1) kl109 larvae, although heterozygous nckap1l lri90 larvae show no observable phenotype without being crossed to Tg(tp1: cdkal1) kl109 . (E and F) Skeletal representation of the intrahepatic biliary network in nckap1l lri90(+/+) ;Tg(tp1:cdkal1) kl109 (E) and nckap1l lri90(+/-) ;Tg(tp1:cdkal1) kl109 (F) larvae computed based on Tg(Tp1-MmHbb:EGFP) um14 expression at 5 dpf. The end points (green), nodes (white), node-node connections (red), and node-end point connections (yellow) are colored separately. Ventral views, anterior to the top. (G-J) Computational analysis-based measurements of the intrahepatic biliary network structures of wild-type (WT) control, nckap1l lri90(+/+) ;Tg(tp1:cdkal1) kl109 and nckap1l lri90 (+/-) ;Tg(tp1:cdkal1) kl109 larvae at 5 dpf. n = 10 for nckap1l lri90(+/+) ;Tg(tp1:cdkal1) kl109 and n = 7 for nckap1l lri90(+/-) ;Tg   6. nckap1l dosage-dependent sensitization to Rac1 inhibitor treatment suggests that Nckap1l acts downstream of Rac1 to regulate intrahepatic biliary network branching morphogenesis. Wild-type and heterozygous nckap1l lri90 mutant larvae were treated with either high dose (50 ug/ml) or low dose (10 ug/ml) Rac1 inhibitor from 3 to 5 dpf, and then the treated larvae were analyzed at 5 dpf. (A-D) Projected confocal images of Tg (Tp1-MmHbb:EGFP) um14 expression in control wild-type (A), high dose (50 ug/ml) Rac1 inhibitor-treated wild-type (B), low dose (10 ug/ml) Rac1 inhibitor-treated heterozygous nckap1l lri90 mutant (C), and low dose (10 ug/ml) Rac1 inhibitor-treated wild-type (D) larvae at 5 dpf. GFP expression in the intrahepatic biliary network is shown in pseudocolored yellow. (E-H) Skeletal representation of the intrahepatic biliary network in wild-type (E), 50 ug/ml Rac1 inhibitor-treated wild-type (F), 10 ug/ml Rac1 inhibitor-treated heterozygous nckap1l lri90 mutant (G), and 10 ug/ml Rac1 inhibitor-treated wild-type (H) larvae computed based on Tg(Tp1-MmHbb:EGFP) um14 expression at 5 dpf. The end points (green), nodes (white), node-node connections (red), and node-end point connections (yellow) are colored separately. Ventral views, anterior to the top. (I-L) Computational analysis-based measurements of the intrahepatic In this study, through a forward genetic approach, we have identified Nckap1l as a novel factor regulating branching morphogenesis of the intrahepatic biliary network. Nckap1l is a component of the WAVE regulatory complex (WRC), which is known to act downstream of Cdk5 and Rac1 to regulate actin dynamics. Consistently, we show that the nckap1l mutation genetically interacts with the Cdk5 (Fig 5) and Rac1 (Fig 6) pathways in biliary epithelial cells, suggesting that we have identified a new signaling branch that acts downstream of Cdk5 and Rac1 to orchestrate branching morphogenesis of the intrahepatic biliary network (Fig 6M).
We initially focused on the nckap1l lri35 mutant phenotype even before identifying the responsible gene because the biliary phenotype calculated by the computational network structure analysis is similar to those of Cdk5-suppressed larvae [8]. It is intriguing that Nckap1l is a component of the WRC, which is directly phosphorylated by Cdk5 in other systems, and we have identified that the nckap1l lri35 mutation actually interacts genetically with the Cdk5 pathway ( Fig 5). These data suggest that precise quantification of mutant phenotypes with our computational analysis could predict the signaling pathway in which the mutated gene lies.
In biliary epithelial cells, Nckap1l exhibits a punctate localization near the plasma membrane ( Fig 3B). The projected confocal images indicated that Nckap1l colocalizes with HSPC300 near the tip of the projecting intrahepatic biliary network (Fig 3K), indicating that the WRC forms near the leading edge of protruding BECs to regulate branching morphogenesis, which is similar to the previous observation that the WRC is known to localize to the leading edges of lamellipodia in the migrating cell [18]. We previously showed the tight correlation between the actin organization in BECs and the intrahepatic biliary network branching pattern [8], in which the consolidated actin organization in BECs correlates with the paucity of intrahepatic biliary network branching. Consistent with this previous observation, we found that actin dynamics in BECs are consolidated in nckap1l lri35 mutant larvae (Fig 3L and 3M), suggesting that changes in BEC actin dynamics might influence the branching pattern of the intrahepatic biliary network.
In this study, we have identified a previously unannotated splice isoform of nckap1l in zebrafish, and the nckap1l lri35 mutation specifically affects this minor splice isoform (Fig 2); this study added a new gene to the list that previously shown to regulate intrahepatic biliary network formation in zebrafish [38][39][40][41][42][43][44][45][46][47][48]. As the nckap1l lri35 mutation does not affect the coding sequence of the major isoform (α) of nckap1l, we assume that the minor splice isoform (ß) is responsible for the intrahepatic biliary network phenotypes observed in nckap1l lri35 mutant larvae. Consistently, injecting the RNA encoding the minor nckap1l isoform at least partially rescued the nckap1l lri35 mutant phenotype (Fig 4M-4O). Although specific binding domains of Nckap1l to other WRC proteins are not precisely determined yet, it will be important to biliary network structures of wild-type, high dose (50 ug/ml) Rac1 inhibitor-treated wild-type, low dose (10 ug/ml) Rac1 inhibitor-treated heterozygous nckap1l lri90 mutant, and low dose (10 ug/ml) Rac1 inhibitor-treated wild-type larvae at 5 dpf. (I) Total network length of the intrahepatic biliary network. (J) Mean network thickness of the intrahepatic network. (K) Network volume (mm 3 ) of the intrahepatic biliary network. (L) Total number of nodes in the intrahepatic biliary network. These data together indicate that low dose (10 ug/ml) Rac1 inhibitor treatment did not cause any observable phenotype in the intrahepatic biliary network of wild-type larvae, but losing one copy of the nckap1l gene significantly enhanced the effect of low dose (10 ug/ml) Rac1 inhibitor treatment and induced a phenotype to closer to that of high dose (50 ug/ml) Rac1 inhibitor treated wild-type larvae. These data suggest that nckap1l acts downstream of Rac1 to regulate intrahepatic biliary network branching morphogenesis. Plots with a shared letter indicate that the difference is not statistically significant. (M) Model of the Cdk5-mediated kinase cascade that regulates branching morphogenesis of the intrahepatic biliary network. We have previously shown that Cdk5 regulates the Pak1/Limk1/Cofilin kinase cascade to regulate actin dynamics. The current study revealed that Cdk5 also regulates the WAVE regulatory complex to regulate branching morphogenesis of the intrahepatic biliary network. In this process, a previously unannotated minor splice isoform of Nckap1l appears to be important to form a functional WAVE regulatory complex in biliary epithelial cells.
https://doi.org/10.1371/journal.pgen.1009402.g006 identify binding partners that specifically bind to the Nckap1l ß specific C-terminus domain (Fig 2D), which is mutated by the lri35 mutation. We detected strong Nckap1l expression in vascular endothelial cells in the liver (Figs 3A and S6), we did not observe overt differences in the vasculature of nckap1l lri35 mutant larvae (S8 Fig). However, in nckap1l lri90 mutant larvae at 6 dpf, although the penetrance is low, we observed disrupted intersegmental vessels (S8 Fig), suggesting that Nckap1 α might be predominantly required for vascular angiogenesis. It is not yet known whether a similar minor splice isoform of NCKAP1L exists in humans; however, the last exon of human NCKAP1L is large. It would be intriguing to clone the short splice form of NCKAP1L in rodents and humans to understand the clinical relevance of this finding. Indeed sequence data derived from patients with biliary atresia and other cholangiopathies should be reexamined focusing on mutations in the last exon of human NCKAP1L after the conventional stop codon, as our data potentially suggest that the genomic region currently recognized as the 3' UTR could encode a previously unannotated minor protein isoform.
Nckap1l knockout mice exhibited a 25-fold increase in the number of circulating neutrophils [16]; however, in zebrafish larvae, the number of neutrophils remains constant in nckap1l lri90 mutant larvae at 5 dpf (S9 Fig), suggesting that nckap1l is not required for the initial differentiation of neutrophils in zebrafish. We did not observe neutrophil accumulation in the liver of nckap1l lri90 mutant larvae at 5 dpf, suggesting that liver inflammation might not happen at this stage. As we found that the nckap1l lri90 mutation genetically interacts with Tg (tp1:cdkal1) kl109 (Fig 5), in which the Cdk5 inhibitor is specifically expressed by biliary epithelial cells within the liver, we assume that nckap1l functions cell-autonomously in biliary epithelial cells to regulate actin remodeling for branching morphogenesis. This view is consistent with our observation that Nckap1l is expressed in biliary epithelial cells (Fig 3B). However, our data do not exclude the possibility that the biliary phenotype seen in nckap1l mutant larvae might be secondary to a phenotype in innate immune cells. We also observed that the nckap1l lri90 mutation genetically interacts with the Rac1 pathway (Fig 6). Since the WRC is known to act downstream of Rac1 [36], we assume that Rac1 regulates the intrahepatic biliary network branching morphogenesis at least in part through the WRC containing Nckap1l. These data suggest that Nckap1l might be working as a signaling hub integrating the Cdk5 and Rac1 pathways to regulate proper branching morphogenesis of the intrahepatic biliary network ( Fig 6M).
In conclusion, we have identified a previously unannotated splice isoform of nckap1l that is necessary for the branching morphogenesis of the intrahepatic biliary network and actin dynamics in biliary epithelial cells. The fact that minor splice isoform of the gene, whose major isoform is known to regulate innate immune cell differentiation and migration, plays pivotal roles in biliary system formation further implies that the correlation between cholangiopathy and inflammation might be due to genes required for both processes.

N-ethylnitrosourea (ENU) mutagenesis and screening procedure
The standard three-generation screen with analysis of mutant phenotypes in F3 larvae was conducted as previously described [52]. In brief, 20 adult male AB/TL zebrafish were exposed to 3 mM ENU at weekly intervals four times. The mutagenized males were crossed to Tg (Tp1-MmHbb:EGFP) um14 fish. Total 228 F2 families representing 332.32 genomes were screened for altered Tg(Tp1-MmHbb:EGFP) um14 expression in the liver at 5 dpf. The phenotype was confirmed and quantified by the custom algorithm [8] as described below. We only collected mutants that showed no overt physical appearance phenotype at 5 dpf. We have recovered 24 alleles, including nckap1l lri35 , showing phenotypes in the intrahepatic biliary network from this screen. The recovered nckap1l lri35 allele was backcrossed to the original AB/TL strain for 9 generations before starting detailed phenotype analyses.

Computational network structure analysis
Confocal z-stack data of Tg(Tp1-MmHbb:EGFP) um14 expression were obtained using a Leica SP5 confocal microscope. The z-step used on the images was 0.25 μm. We used Imaris 8.2 software (Bitplane) to digitally crop the image such that only EGFP expression from the intrahepatic biliary network remained for further analysis. The Liver Analysis Program 5.3 [8] was used for all computational network structure analyses. For all analyses, we confirmed the proper skeletal conversion by overlaying the original confocal image with the converted skeletal image as previously described [8].

Sample preparation, RNA-seq and bioinformatics
Total RNA from 20 wild-type siblings and 20 nckap1l lri35 mutant larvae was collected at 5 dpf using the Qiagen RNeasy mini kit (Qiagen, Cat. 74104) according to the manufacturer's instructions. cDNA for next-generation sequencing was synthesized by using the TruSeq standard total RNA kit (Illumina, 20020598) according to the manufacturer's instructions. Using an Illumina HighSeq 2500, a total of 11.5 and 10.5 millions of paired-end 50-bp reads were obtained from wild-type and mutant cDNA, respectively. Raw reads were aligned to the zebrafish genome (GRCz10) using TopHat [27]. The lri35 mutation was mapped to chromosome 11 by MMAPPR [24]. The NGS and analyzed data are available in the GEO repository at the NCBI (GSE153386).

Nckap1l antibody production
A polyclonal antibody against zebrafish Nckap1l, ab805, was produced by Thermo Fisher Scientific according to their established production protocol. In brief, zebrafish Nckap1l peptides, "RINHIKKCFSDPKRRP", were synthesized, and NZW SPF rabbits were immunized. Serum was collected at day 90 and used at a 1:1 dilution in blocking buffer for immunohistochemistry. This antibody is predicted to recognize both α and ß isoforms of Nckap1l.

Generation of the nckap1l lri90 allele
To create the lri90 mutation, the following crRNA sequence was used: 5'-CUCCGCCAGUUU CAGCUGGUGUUUUAGAGCUAUGCU-3', which was injected into wild-type fertilized eggs. The CRISPR/Cas9 complex was assembled according to the manufacturer's instructions. In brief, 3 μL of 100 μM crRNA, 3 μL of 100 μM Alt-R CRISPR-Cas9 tracrRNA (Integrated DNA Technologies, 209702895), and 94 μL of nuclease free water were mixed. The mixture was then heated to 95˚C. Two microliters of the mixture, 0.5 μL of Alt-R S.P. Cas9 Nuclease V3 (Integrated DNA Technologies, 209702894), 2 μL of phenol red, and 5 μL water were mixed and heated to 37˚C before injection. We injected this solution into fertilized zebrafish eggs, and we subsequently established the nckap1l lri90 allele which disrupts both nckap1l α and ß isoforms.

Liver size measurement
The liver size was measured based on the images of Tg(fabp10:RFP-CAAX) lri2 expression at 5 dpf utilizing the ImageJ software. nckap1l morpholino injection nckap1l splice blocking morpholino (5'-AGCGGCTCCGCTCACCTTCTTGATG-3') was designed and injected as previously described [38]. This morpholino is predicted to block proper splicing of both nckap1l α and ß isoforms.

Genotyping of the lri35 and lri90 alleles
The following primers were used to genotype the lri35 allele: forward primer that we named 17, 5'-GTCAGGATATGCTGGAGATGTG-3'; reverse primer that we named JW-genotypin-g_AR1, 5'-TGATCTGGATTCTGAAGAAGCCACTGA-3'. The lri35 mutation induces an MboII site in the PCR product amplified by these primers. The following primers were used to genotype the lri90 allele: forward primer that we named pam-F1, 5'-GATTAGAGAAAGCTG AGAGCGGAAGTG-3'; reverse primer that we named pam-R1, 5'-ACTGAGGACTTCAGA AGCGGCTCCGCT-3'. The lri90 mutation eliminates the PvuII site in the PCR product amplified by these primers.

PED6-based biliary system functional assay
PED6 treatment and measurement were performed as previously described [8]. In brief, zebrafish larvae obtained from heterozygous nckap1l lri90 mutant fish cross were soaked in PED6-containing media from 4 to 5 dpf, scored PED6 staining in the gallbladder as previously described [8], and genotyped.

Adult liver tissue processing and imaging
12-month-old adult fish were dissected to examine altered Tg(Tp1-MmHbb:EGFP) um14 expression in the liver. The liver lobe fixed in 2% FA in PEM was embedded in 2% GeneMate Low-Melt Agarose (Cat. No. E-3126-25) in PBS. The agarose block was sliced at a thickness of 250 μm with Leica VT1000S vibratome. The slices were scanned on a Leica SP5 confocal microscope. Tg(Tp1-MmHbb:EGFP) um14 expression in the liver was scanned in a z-stack image, and the projected images were generated by Bitplane Imaris software.

Statistics
For pairwise analysis, Student's t-test was used to compare means assuming unequal variance. To compare three or more means, one-way ANOVA followed by Tukey's HSD test was used. In nckap1l lri90 mutant larvae, PED6 transport to the gallbladder is reduced, suggesting that the functionality of the biliary network is impaired. (T and U) Projected confocal images of the adult liver visualized for Tg(Tp1-MmHbb:EGFP) um14 expression in wild-type (T) and nckap1l lri90 mutant (U) 12-month-old fish. The 12.6 um thickness cross sections of the liver were projected and shown. The Tg(Tp1-MmHbb:EGFP) um14 expressing intrahepatic biliary network appears to be thinner and less dense in nckap1l lri90 mutant fish than in wild-type fish. However, as previously reported [46], the variability in adult biliary network phenotype is high. n>10 for WT, and n = 3 for nckap1l lri90 fish. � P<0.05, �� P<0.01, and ��� P<0.001. n.s., not significant. (TIF) S3 Fig. The nckap1l lri90 mutant failed to complement the lri35 mutant, indicating that these two mutations are affecting the same gene. Homozygous nckap1l lri90 mutant fish were crossed to heterozygous lri35 mutant fish, and approximately 50% (20/43) of larvae showed a phenotype indistinguishable from nckap1l lri90 mutant larvae. (A and B) Representative physical appearance of heterozygous nckap1l lri90 (A) and nckap1l lri90/lri35 compound heterozygous (B) larvae at 5 dpf. (C and D) Lateral views of Tg(Tp1-MmHbb:EGFP) um14 expression in heterozygous nckap1l lri90 (C) and nckap1l lri90/lri35 compound heterozygous (D) larvae. In nckap1l lri90/lri35 compound heterozygous larvae, the intrahepatic biliary network is reduced as seen in nckap1l lri90 and lri35 mutant larvae, indicating that these two alleles fail to complement. (TIF) S4 Fig. Sanger sequencing confirms a one-nucleotide insertion in the nckap1l gene. nckap1l cDNA isolated from wild-type (WT) and nckap1l lri35 mutant larvae was sequenced in the 3' to 5' direction. The green box indicates the inserted thymine nucleotide, which induces a frameshift specifically in the minor ß splice isoform of nckap1l. (TIF) S5 Fig. nckap1l expression in wild-type larvae at 5 dpf. (A) qPCR analysis of nckap1l α mRNA expression levels in wild-type control and nckap1l lri35 mutant larvae at 5 dpf. (B) qPCR analysis of nckap1l ß mRNA expression levels in wild-type control and nckap1l lri35 mutant larvae at 5 dpf. The averages of at least three independent experiments are shown. nckap1l ß mRNA expression level is slightly down-regulated while nckap1l α mRNA expression level remains constant in nckap1l lri35 mutant larvae. � P<0.05, n.s., not significant; error bars indicate standard deviation. (C and D) nckap1l mRNA expression in wild-type larvae was examined by in situ hybridization at 5 dpf. Since the entire nckap1l ß gene sequence is part of the ORF and UTR of the nckap1l α gene, we were not able to design the nckap1l ß-specific RNA probe. Instead, we used two different RNA probes for in situ hybridization; one specifically recognizes nckap1l α and the other recognizes both nckap1l α and ß isoforms. (C) nckap1l expression in wild-type larvae at 5 dpf. The RNA probe that recognizes both α and ß isoforms was used. (D) nckap1l expression in wild-type larvae at 5 dpf. The RNA probe that recognizes only the α isoform was used. Black broken lines outline the liver position in C' and D'. nckap1l is expressed widely including in the vertebrate, intestine and swim bladder. In the liver, nckap1l ß appears to be expressed more, as the α and ß isoform probe (C') shows a stronger signal than that of the α isoform specific probe (D'). , and HSPC300 expression levels in wild-type, nckap1l lri90 mutant, and nckap1l lri35 mutant larvae were analyzed by western blotting at 5 dpf. Whole-body homogenates of 5 dpf larvae were used. All WRC protein levels were reduced in nckap1l lri35 mutant larvae, suggesting that WRC proteins were degraded. Tubulin blotting was for loading control. These experiments were repeated three times with similar results. (TIF) S8 Fig. Formation of intersegmental vessels in nckap1l mutant larvae. (A-D) Lateral views of the trunk in wild-type (A), nckap1l lri35 (B), and nckap1l lri90 mutant (C and D) Tg(kdrl: RFP_CAAX) y171 larvae at 6 dpf. In all mutant larvae examined (n = 24 from three independent crosses), we did not observe any overt blood vessel phenotype in nckap1l lri35 mutant larvae at 6 dpf (B). In nckap1l lri90 mutant larvae at 6 dpf, the majority of mutant larvae (n = 27/45 from three independent crosses) show no overt blood vessel phenotype (C). However, we consistently observed disrupted intersegmental vessels (D) in some nckap1l lri90 mutant larvae (n = 18/45). These data suggest that although the phenotype penetrance is low, the nckap1l lri90 mutation can induce a specific morphological phenotype in the blood vessels.