Cftr controls lumen expansion and function of Kupffer’s vesicle in zebrafish

Regulated fluid secretion is crucial for the development and function of many organs in vertebrates, including the kidney, vasculature, brain and ear (Cartwright et al., 2009). During organogenesis, fluid secretion can act as a force driving tubulogenesis. In zebrafish, fluid secretion promotes single lumen formation in the gut (Bagnat et al., 2007) and ventricle inflation in the brain (Lowery and Sive, 2005). Similarly, in mammals, fluid secretion has been shown to be crucial for lung development (Wilson et al., 2007). Loss of fluid regulation can lead to defects in organogenesis. For example, excessive fluid accumulation leads to dramatic expansion of the zebrafish gut lumen (Bagnat et al., 2010) and defects in cilia-dependent fluid clearance can lead to kidney cysts and hydrocephalus (Kramer-Zucker et al., 2005; Moyer et al., 1994; Nauli et al., 2003; Sun et al., 2004).

A major regulator of fluid secretion in vertebrates is the chloride channel cystic fibrosis transmembrane conductance regulator (CFTR). CFTR regulates fluid secretion by controlling the transport of chloride (Anderson et al., 1991), which draws sodium to generate osmotic gradients that drive the movement of water through a tissue. Defects in CFTR function cause cystic fibrosis (CF), a disease in which loss of Cl^- and/or HCO ^-₃ transport impairs fluid secretion and also causes mucus build-up in many organs, including the lungs, intestine and pancreas (Durie and Forstner, 1989; Gaskin et al., 1988; Matsui et al., 1998). The channel is composed of several domains, including twelve transmembrane domains, two nucleotide-binding domains and a unique regulatory domain (R-domain) (Riordan et al., 1989). The R-domain is regulated through phosphorylation by cyclic AMP-dependent protein kinase (PKA) (Berger et al., 1993). Zebrafish Cftr is highly similar to its human ortholog, particularly in domains important for CFTR function (Bagnat et al., 2010). Although CFTR activity has been well characterized in vitro, relatively little is known about its function in vivo, especially during development.

In vertebrates, fluid flow generated by ciliary beating is important for the determination of left-right (LR) asymmetry. In zebrafish, laterality is controlled by a transient, fluid-filled structure called Kupffer’s vesicle (KV) (Essner et al., 2005). KV is functionally homologous to the organs of asymmetry in other vertebrates, including the mouse node and the gastrocoel roof plate in Xenopus (Nonaka et al., 2002; Schweickert et al., 2007). KV develops from a group of dorsal forerunner cells (DFCs) that migrate to the vegetal pole during gastrulation and coalesce to form a fluid-filled spherical structure surrounding a single lumen (Amack et al., 2007; Oteíza et al., 2008). Within the KV lumen, motile cilia drive directional fluid flow leading to asymmetric calcium signaling at the periphery, similar to the mouse node (McGrath et al., 2003; Sarmah et al., 2005). Asymmetric signaling leads to an upregulation of left-sided genes beginning with southpaw (spaw), the zebrafish ortholog of Nodal, a highly conserved signaling molecule required for the specification of LR asymmetry (Brennan et al., 2002; Long et al., 2003; Saijoh et al., 2003). Although it is has been well established that cilia-driven fluid flow is crucial for KV function, the mechanisms that regulate secretion of fluid into KV remain uncharacterized.

Here, we describe a new role for Cftr in the development and function of KV in zebrafish. Using TAL effector nucleases (TALENs), we generated cftr mutants and found that loss of Cftr activity impairs KV lumen expansion and function, causing defects in LR patterning. Using bacterial artificial chromosome (BAC) recombineering we generated a Cftr-GFP transgenic line and observed that cftr is expressed primarily in KV, where the protein localizes apically as the lumen forms. Together, our results demonstrate that Cftr-dependent fluid secretion is crucial for lumen formation and function of KV in zebrafish.

Zebrafish were maintained at 28°C and propagated as previously described (Westerfield, 2000). The following zebrafish lines were used for this work: AB, EK, Tg(sox17:GFP)s870 (Sakaguchi et al., 2006), Tg(fabp10:dsRed, ela:GFP)gz12 (Farooq et al., 2008), TgBAC(cftr-GFP)pd1041, TgBAC(cftr-RFP)pd1042, cftr^pd1048, cftr^pd1049, cftr^pd1050 and Tg(hsp70l:GFP-podxl)pd1080 (this study).

Three TALENs (Miller et al., 2011) were designed to target the sixth exon of cftr using TALEN targeter (Doyle et al., 2012) and constructed using Golden Gate assembly (Cermak et al., 2011). The TALEN used to generate the cftr mutant alleles reported here was composed of the following TAL effector domains: NN NN NN NG NI NG NN NN HD HD HD NI NG NG NG NG NI NG NI NG and NN NG NI HD NI HD NI NN NN NI NG NN HD NI NG NG. Zebrafish were injected into the yolk at the one-cell stage with 100 pg total TALEN RNA and 50 pg of dsRed RNA to mark expressing embryos. Mutant alleles were identified by EcoRV digestion of a PCR product generated with the following primers: cftr-exon6-F, TTGGGCCTAAATTTCAAATGAT; and cftr-exon6-R, TTTGGATGCACAGTAGGCTAA.

A BAC containing cftr (DKEY-270I2) was modified using Red/ET BAC modification plasmids (Genebridges, Heidelberg, Germany). A positive selection cassette for generating C-terminal fusions was developed by constructing a plasmid containing a 20-aa spacer (DLPAEQKLISEEDLDPPVAT), GFP or mRFP-Ruby, an SV40 poly-adenylation sequence, and an FRT-kanamycin-FRT cassette (Lee et al., 2001) into pBluescript. Recombination was performed by amplifying the cassette with the following primers, which contained 50 bp of homology flanking the stop codon: cftr-spGFP-hom-F, CGCAGACCCTGCAAGAGGAGGCAGAGGACAACATCCAGGACACTCGCCTCGATCTCCCCGCCGAACAGAAA and cftr-spGFP-hom-R, TTTAATGTACCATTGGGTGACGGCCTGGGTCACTGAGTCTTTTGGAACGCATTGGAGCTCCACCGCGGTG. The amplicon was then transformed into Red/ET-induced Escherichia coli. The kanamycin was removed from the BAC by expressing Flpase from the p707-Flpe plasmid (Genebridges). The cftr-RFP BAC was further modified by recombining the iTol2-Amp cassette into the loxP site of the pIndigoBAC-536 backbone (Suster et al., 2009). The modified BAC was purified using the Nucleobond BAC-100 Kit (Clontech, Mountain View, CA, USA). The cftr-GFP BAC was linearized using SfiI (NEB) and injected into one-cell-stage embryos. The cftr-RFP BAC was co-injected with 50 pg of transposase into one-cell-stage embryos (Kawakami, 2004; Kwan et al., 2007). Two transgenic lines were established: TgBAC(cftr-GFP)pd1041 and TgBAC(cftr-RFP)pd1042.

The Tg(hsp70l:GFP-podxl)pd1080 line was generated by Gateway recombination with the Tol2Kit (Kwan et al., 2007). GFP-podocalyxin (Meder et al., 2005) was subcloned into pME and assembled with p5E-hsp70l, p3E-polyA and pDestTol2pA2. The resulting plasmid was co-injected into one-cell-stage embryos with 50 pg Transposase RNA. To induce expression, embryos at 50% epiboly were heat-shocked for 30 minutes in a 39°C water bath. GFP-podocalyxin was imaged in conjunction with GFP Counterstain BODIPY TR Methyl Ester dye (Invitrogen).

The probe to detect the cftr transcript by in situ hybridization was PCR amplified from cDNA and ligated into pGEMT-Easy (Promega, Madison, WI, USA) with the following primers: cftr-ish-F, CCAAACCAGACAAAGGCAAA; and cftr-ish-R, GGTGCCATCTCACGATAACTCAA. In situ hybridization was performed as previously described (Marjoram and Wright, 2011; Snelson et al., 2008). Detection of spaw, cmlc2 (myl7 - Zebrafish Information Network), lefty1 and no tail transcripts were performed as previously described (Long et al., 2003; Yelon et al., 1999). The plasmids were linearized and digoxygenin-labeled RNA was generated using the DIG RNA Labeling Kit (Roche). Stained embryos were imaged on a Discovery.V20 stereoscope (Zeiss, Oberkochen, Germany) with an Achromat S 1.0× lens.

Whole-mount immunofluorescence using aPKC (Santa Cruz Biotechnology; 1/100), pan-cadherin (Santa Cruz Biotechnology; 1/1200) and ZO-1 (Invitrogen; 1/500) primary antibodies with goat anti-mouse Alexa568 or goat anti-rabbit Alexa647 secondary antibodies (Molecular Probes; 1/100) was performed as previously described (Li et al., 2011). Acetylated tubulin staining was performed as previously described (Zaghloul and Katsanis, 2011) with the following modifications. Embryos were fixed overnight at 4°C in Dent’s fixative, treated with 10 μg/ml Proteinase K (Sigma) for 1 minute and post-fixed with 4% paraformaldehyde. Mouse anti-acetylated tubulin (Sigma; 1/1000) was detected with goat anti-mouse Alexa568 (Molecular Probes; 1/100). Tailbuds were dissected with a microknife (Fine Science Tools, Foster City, CA, USA), mounted in SlowFade Gold (Invitrogen) and imaged using a Leica SP5 confocal microscope.

Embryos for live confocal imaging were mounted in 4% agarose on slides and immediately imaged on an SP5 confocal microscope (Leica, Wetzlar, Germany) with an HC PL APO 20×/0.70 objective. Imaging of KV flow was performed by injecting fluorescent beads as previously described (Borovina et al., 2010). Differential interference contrast (DIC) microscopy and whole-mount epifluorescence were performed on embryos mounted in 3% methylcellulose and imaged on an Imager M1 (Zeiss) with a EC Plan-Neufluar 10×/0.3 objective.

The TgBAC(cftr-GFP) open reading frame was cloned into pCS2+ with flanking EcoRI sites. RFP was fused to the N-terminus of Clic5b (accession number: BC085448) and cloned into pCS2+ with flanking EcoRI and XhoI sites. Capped RNA was transcribed from NotI linearized plasmid using the mMESSAGE mMACHINE SP6 Kit (Ambion, Grand Island, NY, USA). cftr-GFP (150 pg/embryo) or Arl13b-mCherry (75 pg/embryo) (Borovina et al., 2010) RNA was injected into the yolk of one-cell-stage embryos.

Pharmacological reagents were purchased from Sigma (St Louis, MO, USA). Forskolin was prepared as a 10 mM stock in DMSO and fish were treated in egg water at 10 μM. IBMX (3-isobutyl-1-methylxanthine) was prepared as a 100 mM stock in DMSO and embryos were treated at 40 μM in egg water. Ouabain was prepared as a 1 mM stock in DMSO and embryos were treated at 1 μM in egg water. Embryos were treated from 50% epiboly until they were imaged at the 10- to 12-somite stage (ss).

Human HEK293 and Cos-7 cells were cultured in DMEM with 10% fetal bovine serum and 1% penicillin-streptomycin (Invitrogen). Cells were transfected on glass coverslips with either pCDNA-Cftr-GFP or pCDNA-Cftr^pd1048-GFP using Lipofectamine 2000 (Invitrogen) and fixed the following day in 4% paraformaldehyde. Fixed cells were stained with DAPI and imaged on an Imager M1 (Zeiss).

Measurements of KV lumen area and cilia length were performed using ImageJ (NIH, Bethesda, MD, USA) and analyzed for statistical significance using Student’s t-test in Prism (Graphpad, La Jolla, CA, USA). Comparisons between KV phenotype and organ laterality were performed using a χ² test in Prism (Graphpad).

To investigate the role of Cftr during zebrafish development we generated cftr mutants using TALENs (Cermak et al., 2011; Huang et al., 2011; Miller et al., 2011). Three TALEN pairs for cftr were constructed, transiently expressed in zebrafish embryos and then screened for activity. Embryos injected with the TALEN pair showing the highest transient activity were raised to establish mutant lines. The TALEN was targeted to the EcoRV site in the sixth of 27 cftr exons, corresponding to the third transmembrane domain. This exon precedes several domains crucial for Cftr function, including the chloride pore, regulatory domain and both nucleotide-binding domains (Fig. 1A,B). After screening ten TALEN-injected fish for disruption of the EcoRV restriction site (Fig. 1C), we identified three cftr alleles, including a two-amino acid deletion, cftr^pd1048, and two frameshift mutations, cftr^pd1049 and cftr^pd1050 (Fig. 1D). The cftr^pd1049 allele is predicted to code for 56 incorrect amino acids past the lesion before encountering a stop codon (Fig. 1E), whereas cftr^pd1050 generates a stop codon at the site of the lesion. To characterize the cftr^pd1048 mutation, we cloned and expressed GFP-tagged wild-type (WT) Cftr (Cftr-GFP) and Cftr^pd1048-GFP in Cos-7 cells. Unlike WT Cftr-GFP, Cftr^pd1048-GFP was largely absent at the cell surface and localized mostly to intracellular membranes resembling the endoplasmic reticulum, suggesting that the shortened transmembrane domain encoded by the mutant allele affects biosynthetic transport (supplementary material Fig. S1A,B). Accordingly, in transfected cells Cftr^pd1048 lacked the mature, fully glycosylated form of the protein, as judged by western blot analysis (supplementary material Fig. S1C).

To analyze Cftr function during development, the cftr mutations were crossed into the Tg(fabp10:dsRed, ela:GFP)gz12 background, which expresses dsRed in the liver and GFP in the exocrine pancreas (Farooq et al., 2008). While examining homozygous mutant embryos, we found inversion of liver and pancreas situs. In zebrafish, the liver normally develops on the left side of the abdomen; however, in 27% of cftr^pd1049 mutants the liver developed on the right (n=30 mutants), compared with 0% of their WT siblings (n=103 WT) (Fig. 2A,B).

To test whether the anatomical positioning of other organs is also affected in cftr mutants, we examined heart looping, one of the earliest morphological indicators of organ laterality. The heart, marked by cmlc2 expression, normally loops to the left. In ∼31% of cftr mutant embryos the heart looped to the right (n=138 mutants), in contrast to only 1% of WT siblings (n=408 WT) (Fig. 2C,D). Homozygous mutants survived to adulthood and the females were moderately fertile, allowing the generation of maternal zygotic cftr^pd1049 mutants. Maternal zygotic cftr^pd1049 fish were morphologically identical to zygotic cftr^pd1049 and had similar rates of reversed heart looping (31%, n=26). Defects in liver and heart orientation were primarily concordant and only a few cases of heterotaxia were observed.

To investigate early events in LR patterning, we examined the expression of spaw, a gene asymmetrically expressed in the left lateral plate mesoderm of the embryo at 20 ss and an important LR patterning determinant (Long et al., 2003). Whereas 99.3% of WT siblings expressed spaw exclusively on the left (n=146), cftr^pd1049 mutants displayed a range of spaw expression phenotypes including exclusively left (30.6%), left dominant (25.0%), bilateral (23.6%), right dominant (9.7%) and exclusively right (11.1%) (n=73) (Fig. 2E,F). These data indicate that cftr is required for the establishment of LR asymmetry in zebrafish before the onset of spaw expression.

Aberrant spaw expression and organ laterality can be caused by defects in processes that establish and restrict asymmetric signaling. Dorsal midline structures, such as the notochord and floorplate, function as a barrier to restrict spaw to the left side of the embryo and loss of barrier function has been shown to result in bilateral spaw expression (Long et al., 2003). Additionally, midline expression of lefty1, a spaw antagonist, is required for left-sided restriction of spaw expression (Bisgrove et al., 1999). To determine whether cftr^pd1049 mutants have midline barrier defects, we examined the expression of midline markers as well as the gross morphology of WT and mutant embryos. At 22 ss, expression of the notochord marker no tail (ntl) in cftr^pd1049 mutants was indistinguishable from that of WT siblings (supplementary material Fig. S2A,B). We also examined the expression of lefty1 and found that in cftr^pd1049 mutant embryos lefty1 expression was similar to that of WT siblings (supplementary material Fig. S2C,D). In addition, the notochord and floorplate of cftr^pd1049 mutants appeared to be completely intact at 24 hours post-fertilization (hpf) as judged by DIC microscopy (supplementary material Fig. S2E,F). Brightfield whole-mount imaging of cftr^pd1049 mutants at 24 and 48 hpf showed no obvious morphological defects (supplementary material Fig. S2G-J). Thus, defects in spaw expression and LR patterning in cftr^pd1049 mutants are probably not due to defects in midline integrity.

To understand better how Cftr functions in LR asymmetry, we next investigated the development of KV, a transient, fluid-filled organ important for the specification of LR asymmetry in zebrafish (Essner et al., 2005). Examination of cftr mutants by DIC microscopy revealed that the KV lumen was absent in the frameshift alleles cftr^pd1049 and cftr^pd1050 (Fig. 3A-D). This phenotype is fully penetrant; all homozygous cftr mutants displayed defects in KV lumen morphogenesis (n=72). In cftr^pd1048, the lumen was present, but severely reduced in size (Fig. 3E,F), suggesting that this may be a hypomorphic allele.

To investigate the morphology of KV in greater detail, we crossed cftr mutants into the Tg(sox17:GFP)s870 background, a well-established marker of DFCs and KV (Oteíza et al., 2008; Sakaguchi et al., 2006). To visualize cilia, we injected RNA encoding Arl13b-mCherry into one-cell-stage embryos (Borovina et al., 2010). Using live confocal imaging, we found that in cftr^pd1049 mutants the DFCs migrated and clustered to form a structure similar in size to the WT KV. However, in the mutants, lumen inflation did not occur and resulted in a central plate of ciliated, sox17:GFP-positive cells with no discernible luminal space (Fig. 3G,H).

Next, we examined KV development in the hypomorphic cftr^pd1048 allele. Using live confocal imaging we observed that, at 10 ss, mutants homozygous for this allele had a KV with a very small central lumen (∼30% of the area compared with WT siblings) containing Arl13b-mCherry-positive cilia that appeared motile (Fig. 3I,J). Owing to line averaging during confocal microscopy, motile cilia appear as a fan-shaped blur (Borovina et al., 2010). At 12 ss, in some rare cases, the small lumen inflated to a size large enough to reveal seemingly normal, but highly crowded motile cilia (supplementary material Fig. S3A-D), suggesting that Cftr function is not required for cilia morphogenesis or motility in KV. We then investigated whether these mutants formed a lumen large enough to allow for proper specification of LR asymmetry. We examined heart looping and liver orientation in cftr^pd1048 mutants and found that they had a similar rate of LR asymmetry defects compared to the null (cftr^pd1049) allele (data not shown). To determine whether fluid flow is normal in the hypomorphic allele, we injected fluorescent beads into the KV lumen. In the cftr^pd1048 KV, fluid flow appeared turbulent, indicating the organ does not function normally when fluid secretion and lumen size are significantly reduced (supplementary material Fig. S3E,F).

To examine KV ciliogenesis further in the absence of Cftr function and lumen expansion, we determined the number and average length of cilia in KV. We found no significant difference in either the number or length of cilia between cftr^pd1049 mutants and their WT siblings (Fig. 3K-N), indicating that the volume of the KV lumen does not regulate ciliogenesis.

We next examined the apical-basal polarity of KV in cftr^pd1049 mutants, by investigating the localization of several polarity markers. We characterized membrane polarity by examining the localization of the tight-junction protein ZO-1 and the basolateral marker, Cadherin. In mutant embryos, ZO-1 and Cadherin were properly localized in the absence of lumen expansion (Fig. 4A,B). A higher magnification view of the KV epithelium shows that ZO-1 appears completely apical and Cadherin is absent from the apical membrane in WT and cftr^pd1049 mutant embryos (Fig. 4A′-B″′). We next examined aPKC, a peripheral membrane protein localized to the apical membrane in KV (Amack et al., 2007). In cftr^pd1049 mutants, we observed a plate of aPKC-positive membrane on the center (apical side) of the sox17:GFP-positive cluster (Fig. 4C,D). We also generated a transgenic line expressing an integral membrane apical marker, GFP-podocalyxin (GFP-podxl) (Meder et al., 2005) under the control of a heat shock promoter. We observed that in cftr^pd1049 mutants, at 3 ss, GFP-podxl was also localized in a central plate of apical membrane in KV (Fig. 4E,F). To observe the morphology of the KV lumen in cftr^pd1049 mutants better, we live-imaged the localization of an apical peripheral membrane protein, RFP-Clic5b, in Tg(sox17:GFP)-expressing embryos. Injection of RFP-clic5b RNA marked the apical membrane in WT and cftr^pd1049 mutant KV (Fig. 4G,H). Thus, apical-basal polarity in the KV epithelium develops properly in cftr mutants. Altogether, these data indicate that cftr is crucial for lumen expansion but not ciliogenesis or apical-basal polarization during KV morphogenesis.

To determine where Cftr functions during KV lumen formation, we examined cftr expression using in situ hybridization and live imaging. By in situ hybridization, cftr expression was highly enriched in KV at 3 ss, a stage when the lumen is expanding (Fig. 5A). By 10 ss, the transcript appeared to be downregulated and cftr expression was also observed in the chordamesoderm (Fig. 5B).

To understand the dynamics of Cftr expression and localization better in live embryos, we used recombineering to generate a Cftr-GFP fusion protein. The cftr BAC (DKEY-270I2) used contains ∼50 kb of genomic DNA upstream and 100 kb downstream of the coding sequence, and is likely to include critical regulatory information (Fig. 5C). To generate a C-terminal fusion protein, we replaced the stop codon of cftr with GFP, separated by a sequence encoding a 20 amino acid spacer to provide some insulation from GFP. C-terminal fusion proteins of human CFTR maintain similar localization and channel activity to untagged CFTR (Benharouga et al., 2003). We then repeated the recombineering procedure to generate an RFP fusion and established two transgenic lines, TgBAC(cftr-GFP)pd1041 and TgBAC(cftr-RFP)pd1042, that have identical expression patterns at all stages observed. At 10 ss, by whole-mount epifluorescence, Cftr-GFP was highly restricted to KV (Fig. 5D-G). Next, we performed live, time-lapse imaging of TgBAC(cftr-RFP) in conjunction with cytosolic Tg(sox17:GFP) and found that Cftr-RFP was localized apically in KV by 1 ss and throughout the initial stages of lumen formation as multiple small lumens coalesced into a single lumen (Fig. 6A,B; supplementary material Movie 1). By 10 ss, Cftr-RFP remained apically localized as the lumen continued to expand (Fig. 6C,D). At 15 ss, prior to KV disassembly, Cftr-GFP expressed from TgBAC(cftr-GFP) remained localized to the apical membrane (Fig. 6E,F). Together, these data indicate that Cftr is expressed and apically localized in KV throughout its morphogenesis.

Next, we tested whether modulation of Cftr channel activity could regulate the luminal volume of KV. Cftr is strongly activated by phosphorylation of its R-domain by PKA (Berger et al., 1993). To activate Cftr, we treated fish with a cocktail of forskolin and IBMX. These drugs synergistically elevate cAMP levels, rapidly increasing PKA activity, and result in a potent activation of Cftr. PKA activation from 50% epiboly to 12 ss led to a 66% increase in the area of the KV lumen (Fig. 7A-C). We also tested whether the KV lumen could be reduced in size by inhibiting fluid secretion. The ion gradients driving Cftr-dependent fluid secretion are generated by the Na⁺/K⁺-ATPase. Treatment with low concentrations of ouabain, a potent and specific inhibitor of the Na⁺/K⁺-ATPase, from 50% epiboly to 10 ss decreased the area of the KV lumen by 33% (Fig. 7D-F). We further investigated the morphology of KV in fish treated with activators or inhibitors of fluid secretion using live confocal imaging and found that, although the lumen size was changed, the structure was otherwise organized properly, including the development of motile cilia (Fig. 7G-I).

We next tested whether Cftr function is required for fluid secretion specifically in KV by rescuing lumen expansion defects in several ways. We began by injecting WT and cftr mutant embryos with RNA encoding a Cftr-GFP fusion. Because we used the KV phenotype to distinguish mutant from WT embryos, we assayed KV rescue by examining the percentage of phenotypically mutant KV in whole clutches. A cross between parents heterozygous for cftr^pd1049 resulted in a failure of KV lumen inflation in 25% of the clutch. Injection of cftr-GFP RNA at the one-cell stage reduced the proportion of the clutch that failed to undergo KV lumen expansion to ∼10%, indicating that homozygous cftr^pd1049 mutants were partially (∼60%) rescued by RNA injection at the one-cell stage (Fig. 8A-D). The failure to completely rescue KV lumen expansion was probably due to mosaicism of the injected RNA (Carmany-Rampey and Moens, 2006). To test whether cftr is required in KV, we crossed cftr^pd1049 mutants to the TgBAC(cftr-GFP) line, which is almost exclusively expressed in KV during lumen morphogenesis. This transgene was able to completely rescue lumen expansion, whereas non-expressing siblings maintained a 25% failure of KV lumen expansion (n=289) (Fig. 8E). Additionally, all TgBAC(cftr-GFP)-expressing fish had normal heart looping, indicating that the transgene was also able to rescue organ laterality. The TgBAC(cftr-GFP) rescue suggests that Cftr function is required specifically in KV and that the BAC transgene encodes a functional fusion protein.

We also attempted to phenocopy the cftr^pd1049 KV phenotype by injecting a morpholino against Cftr (Bagnat et al., 2010) into DFCs (Amack and Yost, 2004). We found that DFC-specific injections were unable to prevent KV lumen expansion owing to mosaic uptake of the morpholino (Amack and Yost, 2004) (data not shown). This is not surprising given that fluid secretion is expected to function non-cell-autonomously.

Next, we determined whether forskolin and IBMX were acting through Cftr in KV by treating WT and cftr^pd1049 mutants with these activators of fluid secretion. Forskolin and IBMX treatment failed to rescue KV lumen expansion in cftr^pd1049 mutants, indicating that cAMP-stimulated fluid secretion acts through Cftr (Fig. 8F). Taken together, these studies demonstrate that in zebrafish, Cftr functions in KV to drive fluid secretion crucial for lumen expansion and morphogenesis of KV and LR patterning of the embryo.

Here, we describe a new role for Cftr in the regulation of fluid secretion necessary for KV morphogenesis and function. Live imaging of KV lumen morphogenesis showed several small lumens that coalesce into a single, central lumen. In the absence of Cftr activity, the KV lumen fails to inflate, indicating that fluid acts as a force driving lumen expansion and might also promote lumen coalescence. The zebrafish gut undergoes a similar process of de novo lumen formation, beginning with inflation of multiple small lumens, followed by their coalescence into one (Bagnat et al., 2007) and a similar role in de novo lumen formation and expansion has been shown for apical membrane or secreted mucins during tubulogenesis in the vertebrate vasculature (Strilić et al., 2010) and in the ommatidium (Husain et al., 2006) and hindgut (Syed et al., 2012) in Drosophila. The role of luminal content as a driving force during tube formation is most clearly exemplified by the Drosophila tracheal system, which becomes filled with a solid chitin matrix, then liquid and finally gas during its development (Tsarouhas et al., 2007). Altogether, these studies signify that the filling of the lumen is crucial for tubulogenesis in metazoans.

Although complete loss of Cftr function abrogated inflation of the KV lumen, it did not affect the specification, migration or the clustering of DFCs. Ciliogenesis has previously been shown to proceed coordinately with KV lumen expansion, raising the possibility of crosstalk between luminal volume and cilia length (Oteíza et al., 2010). Here, we found that in the absence of fluid secretion into the KV lumen, cilia morphogenesis and motility were unaffected, indicating that fluid secretion and flow are specified largely independently.

Although KV cilia length and number were unchanged in cftr^pd1049 mutants, the absence of a KV lumen resulted in bilateral spaw expression. Previous research has demonstrated that bilateral spaw expression can be caused by defects in the midline barrier, which functions to prevent the diffusion of Spaw to the right side of the embryo (Long et al., 2003). In cftr^pd1049 mutants, proper midline expression of ntl and lefty1 indicate that midline integrity is not perturbed. Moreover, mutations that affect midline integrity result in heterotaxia, a phenotype rarely observed in cftr^pd1049 mutants. This difference in organ situs in cftr^pd1049 mutants can be explained by the fact that a majority of embryos with bilateral spaw expression did not show equivalent left and right spaw, but rather that expression on one side was more pronounced and extended anteriorly. Previous research in Xenopus has demonstrated that in embryos with bilateral Nodal, in which one side displays dominant expression, organ position is specified concordantly (Ohi and Wright, 2007). The remaining pool of cftr^pd1049 embryos with equivalent left- and right-sided spaw expression might have been left-biased in a way in situ hybridization was not sensitive enough to detect. Similar patterns of bilateral spaw expression have also been observed upon loss of function of the Spaw antagonist Charon (Dand5 - Zebrafish Information Network), which also has no associated midline defect (Hashimoto et al., 2004). Although we are unable to completely rule out a contribution from the midline, our data strongly suggest that cftr is required specifically in KV for proper specification of organ laterality. However, the fact that we observe defects in organ laterality in only ∼30% of cftr^pd1049 mutants, instead of a complete randomization of laterality (50% penetrance), suggests that fluid flow in KV is not the only source of asymmetric information in the early zebrafish embryo. Consistent with this interpretation, previous studies in chicken, frog and zebrafish (Kawakami et al., 2005; Levin et al., 2002) suggested that KV might function to amplify earlier asymmetries, such as differences in H⁺/K⁺-ATPase activity and thus bias organ laterality in the absence of KV function.

Although our results demonstrate that Cftr plays an important role in the development of LR asymmetry in the zebrafish, it should be noted that organ laterality defects have not been observed in cystic fibrosis (CF) patients. This is probably due to morphological differences between the teleost KV and the laterality organs in other vertebrates. Whereas KV in zebrafish is an enclosed structure, the node in mammals and the gastrocoel roof plate in Xenopus are indentations, where fluid may freely diffuse into the organ (Blum et al., 2007; Shook et al., 2004; Sulik et al., 1994). In fact, passing artificially generated fluid flow over the mouse node is sufficient to reverse organ laterality (Nonaka et al., 2002). It is also unknown whether cftr orthologs are expressed in the mouse node or frog gastrocoel roof plate. Further investigation will be required to determine whether embryonic fluid secretion is specifically required in the laterality organs of other vertebrates or if they utilize fluid secreted from other sources.

In CF patients, reduced fluid secretion leads to mucus buildup in the lungs, liver and pancreas, disrupting their function (Durie and Forstner, 1989; Gaskin et al., 1988; Matsui et al., 1998). The recent development of new animal models of CF will greatly improve understanding of the pathophysiology of CF (Rogers et al., 2008; Sun et al., 2010). Clinical data from patients and new animal models, such as the pig and the ferret, suggest that some defects associated with CF might arise during development (Imrie et al., 1979; Olivier et al., 2012; Sturgess, 1984), a stage that remains difficult to access in these models with limited genetic tools. A developmentally accessible model will provide valuable insights into CFTR regulation and CF pathophysiology in various organs. The development of zebrafish cftr mutants provides an in vivo model in which to study human CFTR and assay drugs designed to correct the ΔF508-CFTR mutation and other pathologically relevant mutations. Future work may also shed light into the role Cftr plays in the development of the pancreas and other organs.

In summary, this study identifies a new role for Cftr in the regulation of fluid secretion into KV and for the development of LR asymmetry. It also highlights the importance of fluid secretion in lumen expansion during vertebrate morphogenesis. The relative simplicity and experimental accessibility of KV compared with other organs undergoing de novo lumen formation make KV an attractive model for studying fundamental mechanisms of lumen formation in vertebrates.

We thank Sean Ryan for help with cell culture experiments; members of the Bagnat lab for helpful discussions throughout this project; members of Ken Poss’s lab for discussions and assistance with BAC recombineering; James Patton and Joshua Gamse for in situ hybridization probes; Brian Ciruna for an Arl13b-mCherry expression plasmid; Didier Stainier for providing the Tg(sox17:GFP) line; and Cagla Eroglu, Terry Lechler and Ken Poss for critical reading of this manuscript.