Lethal giant larvae 2 regulates development of the ciliated organ Kupffer’s vesicle

In several organs, motile cilia extending from epithelial cells beat in a coordinated fashion to move fluids. Disruption of proteins that regulate cilia function can result in a broad spectrum of clinical disorders known as ciliopathies (Sharma et al., 2008). In the zebrafish embryo, Kupffer’s vesicle, otic vesicles and the pronephric ducts are examples of organs that assemble motile cilia that project into a lumen and generate fluid flow that is necessary for normal development (Essner et al., 2005; Kramer-Zucker et al., 2005; Yu et al., 2011). In ciliated organs, formation and maintenance of cilia requires docking of a basal body to the apical cell surface, vesicle trafficking of ciliary proteins from the Golgi apparatus to the cilium and movement of cargos through the cilia via intraflagellar transport (Ishikawa and Marshall, 2011). At the same time, formation of a lumen requires specific modification of the cytoskeleton, polarized vesicle trafficking to the lumenal surface and regulation of cell junctions (Rodríguez-Fraticelli et al., 2011). Our understanding of how ciliogenesis and lumenogenesis are regulated as individual processes remains incomplete, but even less is known about the relationship between these two processes. Here, we show that zebrafish Lethal giant larvae 2 (Lgl2; Llgl2 - Zebrafish Information Network) is a regulator of both lumenogenesis and ciliogenesis and that these processes can be uncoupled in developing ciliated organs.

First identified as a tumor suppressor in Drosophila (Gateff, 1978), Lgl proteins regulate several cellular activities, including apicobasal polarization, asymmetric division, and migration (Atwood and Prehoda, 2009; Betschinger et al., 2003; Kaplan et al., 2009; Tanentzapf and Tepass, 2003). In epithelial cells, Lgl interacts antagonistically with atypical protein kinase C (aPKC) to establish basolateral membrane domains and to position cell junctions (Betschinger et al., 2005; Plant et al., 2003). Lgl may target proteins to specific membrane locations by regulating vesicle trafficking. Lgl homologs in yeast (Sro7p and Sro77p) and mammals (Lgl1 and Lgl2) bind membrane t-SNAREs (target soluble N-ethylmaleimide attachment protein receptors) that mediate fusion of post-Golgi transport vesicles to target membranes (Lehman et al., 1999; Müsch et al., 2002), suggesting a link between Lgl and polarized exocytosis. Sro7p also interacts with the exocyst complex (Zhang et al., 2005), which tethers vesicles to the plasma membrane, and Sec4p (Grosshans et al., 2006), a member of the Rab family of GTPases that regulate vesicle trafficking and exocyst function. Recently, mammalian Lgl1 was shown to activate Rab10 to control vesicle trafficking during neuronal axon growth (Wang et al., 2011b).

Despite the involvement of Lgl proteins in crucial cellular processes, functions for Lgl during early vertebrate embryonic development remain largely unknown. Although Lgl1 and Lgl2 are widely expressed in the mouse embryo (Klezovitch et al., 2004), knockout of Lgl1 (Llgl1 - Mouse Genome Informatics) appears to specifically disrupt brain development and knockout of Lgl2 (Llgl2 - Mouse Genome Informatics) only affects placental morphogenesis. Lgl1^-/- mice show defects in cell polarity and apical-junctional complex formation in neuronal cells and die neonatally from severe hydrocephalus (Klezovitch et al., 2004), whereas Lgl2^-/- mice are born small, but develop into normal adults (Sripathy et al., 2011). In zebrafish, a loss-of-function lgl2 mutant (penner) develops epidermal defects at 4 days post-fertilization due to an absence of hemidesmosomes and hyperproliferation in the basal epidermis (Sonawane et al., 2005), and co-knockdown studies have shown that Lgl1 and Lgl2 redundantly control the epithelial organization of the lateral line organ (Hava et al., 2009). These tissue-specific and/or late-appearing phenotypes suggest that some Lgl functions during embryogenesis are masked by redundancy or maternal expression.

In this study, depletion of both maternal and zygotic Lgl2 expression in the early zebrafish embryo has uncovered roles for Lgl2 in the development of ciliated epithelia in Kupffer’s vesicle (KV), as well as in otic vesicles and pronephric ducts. Analyses of KV morphogenesis revealed that loss of Lgl2 function disrupted formation of both the cilia and the lumen. Consistent with these defects, left-right patterning was disrupted in Lgl2-depleted embryos, identifying a role for Lgl2 in left-right axis specification. Focusing on the role of Lgl2 in lumen formation, we found that Lgl2 genetically interacted with the GTPase Rab11a to regulate E-cadherin (also known as Cadherin 1, epithelial) at cell-cell contacts. These results identify new functions for Lgl proteins in the vertebrate embryo and implicate Lgl2 in polarized vesicle trafficking, which is essential for development of organs with motile cilia.

Wild-type TAB zebrafish (Danio rerio) and transgenic Tg(sox17:GFP) (Sakaguchi et al., 2006) and Tg(cmlc2:GFP) (Huang et al., 2003) zebrafish were obtained from the Zebrafish International Resource Center (ZIRC). Tg(shh:GFP) zebrafish (Haga et al., 2009) were a kind gift from Shao Jun Du. Additional transgenic lines were Tg(dusp6:memGFP)^pt19 (Wang et al., 2011a) and Tg(β;-actin:Lyn-TdTomato) (J.C., unpublished). Embryos were staged as described (Kimmel et al., 1995).

PCR-amplified lgl2 cDNA was inserted into the pCRII TOPO vector. lgl2 and previously described sox17, spaw, lft1 and shha cDNA constructs (Alexander and Stainier, 1999; Long et al., 2003; Bisgrove et al., 1999; Krauss et al., 1993) were used to generate in vitro-synthesized RNA probes labeled with digoxygenin (Roche DIG RNA Labeling Kit). In situ RNA hybridizations were performed as described (Gao et al., 2010).

Morpholino oligonucleotides (MOs) were obtained from Gene Tools, LLC. Two previously described MOs were used to block Lgl2 translation: Lgl2 MO-1 (5′-GCCCATGACGCCTGAACCTCTTCAT-3′) (Sonawane et al., 2009) and Lgl2 MO-2 (5′-AGCCGGGACTCAAACTGCCCTCTCT-3′) (Hava et al., 2009). Rab11a MO (5′- GTATTCGTCGTCTCGTGTCCCCAT-3′) has also been previously characterized (Westlake et al., 2011). MOs were injected between the 1- and 2-cell stages to allow the MO to be distributed to all embryonic cells. Optimal MO doses used (unless otherwise stated) were: 4.4 ng of Lgl2 MO-1, 2.5 ng of Lgl2 MO-2 and 2 ng of Rab11a MO. All MOs were co-injected with 4 ng p53 MO (5′-GCGCCATTGCTTTGCAAGAATTG-3′) to mitigate off-target effects (Robu et al., 2007). Sub-optimal MO doses used were: Lgl2 MO-1^low=1.7 ng and Rab11a MO^low=0.8 ng. Co-injection of these two MOs at their sub-optimal levels was performed at 1- to 2-cell stages. To conduct rescue experiments, full-length zebrafish lgl2 was PCR amplified from a zebrafish cDNA library and cloned into pCS2+ vector. Seven bases were changed in the MO binding site to make the mRNA MO resistant. This cDNA was used to synthesize capped mRNA using mMessage mMachine (Ambion) that was purified using Bio-Gel P-6 Micro Bio Spin columns (BioRad). Titration of synthetic capped lgl2 mRNA was performed to determine the concentration that resulted in minimal embryo defects. For Lgl2 rescue experiments, Lgl2 MO-1 was co-injected with 150 pg lgl2 mRNA or 150 pg mRNA encoding an Lgl2:EYFP fusion protein (see below), and Lgl2 MO-2 was co-injected with 100 pg lgl2 mRNA. For Rab11a rescue experiments, zebrafish rab11a with five base changes in the MO binding site was amplified from cDNA and cloned into pCS2+ vector. In vitro synthesized mRNA (mMessage mMachine; 100 pg) was co-injected with Rab11a MO at the 1-cell stage for rescue experiments. All MO and rescue data are pooled from at least three independent experiments. To generate an Lgl2:EYFP fusion construct, the protein coding sequence of lgl2 was inserted upstream of EYFP in the PCS2+ vector using ClaI and XhoI restriction sites. The lgl2 stop codon was mutated to cysteine using the QuikChange Site-directed Mutagenesis Kit (Stratagene) to create the fusion protein. To visualize fusion protein localization, 200 pg of in vitro synthesized mRNA (mMessage mMachine) was injected into embryos.

Primary antibodies used were: mouse anti-acetylated tubulin (1:200, Sigma), rabbit anti-aPKC (1:100, Santa Cruz), mouse anti-E-cadherin (1:200, BD Transduction Laboratories), rabbit anti-Lgl2 (1:500, a kind gift from Dr Mahendra Sonawane) (Sonawane et al., 2009), mouse anti-ZO-1 (1:200, Invitrogen) and rabbit anti-GFP (1:200, Molecular Probes). Embryos were fixed in 4% paraformaldehyde in PBS with 0.25% Triton X-100 at 4°C overnight and then de-chorionated in 1× PBS. Embryos were permeabilized in blocking solution containing 1× PBS, 0.5% Triton X-100 and 5% goat serum for 4 hours. Primary antibodies were diluted in fresh blocking solution and incubated with embryos at 4°C overnight and then washed in PBS with 0.5% Triton X-100 (three 30-minute washes) at room temperature. AlexaFluor 488-, 568- and 647-conjugated anti-rabbit and anti-mouse secondary antibodies (Molecular Probes) were used at 1:200 in 5% goat serum. AlexaFluor 488 and Rhodamine Phalloidin (Molecular Probes) were each diluted at 1:200 from stock and incubated together with the secondary antibody at 4°C overnight. The stained embryos were then washed in PBS with 0.5% Triton X-100 (three 30-minute washes) at room temperature. Image acquisition was performed using a Perkin-Elmer UltraVIEW Vox spinning disk confocal microscope or an SP5 Leica laser scanning confocal microscope. Image analyses, including quantification of KV cell number (GFP), KV cilia number and length (acetylated tubulin) and KV lumen area at the middle focal plane of KV (largest KV area based on aPKC staining), were performed using ImageJ software. ImageJ was also used to quantify E-cadherin staining at cell junctions. The mean gray level (per pixel) was determined along continuous lateral membranes (adherens junctions) >5 μm starting from the apical region of KV cells. This mean gray level (fluorescence intensity) was normalized to 1 with reference to the MO control embryos.

Double-transgenic Tg(sox17:GFP; β;-actin:Lyn-TdTomato) embryos were generated by crossing Tg(sox17:GFP) and Tg(β;-actin:Lyn-TdTomato) parents. Uninjected control embryos or embryos injected with 4 ng Lgl2 MO-2 were mounted in 0.7% low melting temperature agarose at the 1-somite stage and then imaged on a Lavision Biotec TrimScope II 2-photon with optical parametric oscillator (OPO) using a Zeiss 20× 1.0 water dipping lens (Olympus BX51 WI stand). Live embryos were maintained at 28.5°C and images were captured every 450 seconds. Four-dimensional datasets were processed using Imaris software (Bitplane) to correct for KV drift. The coronal plane was exported and assembled into a movie after contrast adjustment. KV lumen volume was quantified at the 8-somite stage using Imaris software. Taking a parallel approach, KV was imaged at the 8-somite stage in live Tg(dusp6:memGFP) embryos using spinning disk confocal (Perkin Elmer) microscopy. ImageJ software was used to create a contrasting region of interest (ROI) corresponding to the KV lumen. This ROI was imported into Volocity software (Perkin Elmer) to measure its volume.

All statistical analyses were carried out using one-factor ANOVA and Fisher’s least significant difference post-hoc test. P-values <0.05 were considered to be statistically significant.

The zebrafish penner mutant has revealed roles for Lgl2 in the epidermis at 4-5 days post-fertilization (dpf) (Reischauer et al., 2009; Sonawane et al., 2005; Sonawane et al., 2009), but the pronounced expression of Lgl2 during earlier zebrafish development suggested additional roles for this protein. RNA in situ hybridization analysis showed that lgl2 is maternally supplied (Fig. 1A,B) as previously reported (Sonawane et al., 2005). At the 80% epiboly stage [∼8 hours post-fertilization (hpf)], lgl2 expression was detected in cells of the enveloping layer and was strongly expressed in dorsal forerunner cells (DFCs) that give rise to KV (Cooper and D’Amico, 1996; Melby et al., 1996) (Fig. 1C). At the 2- to 3-somite stages (∼11 hpf), lgl2 was prominently expressed in ciliated KV cells (Fig. 1D,E). At 24 hpf, lgl2 showed a basal level of expression, with higher expression in organs that contain motile cilia, including nasal placodes, otic vesicles and pronephric ducts (Fig. 1F-H). These analyses suggest that Lgl2 functions in the development of ciliated organs.

To determine the role of Lgl2 during early zebrafish development, we used two previously described antisense morpholino oligonucleotides (MOs) designed to block translation of lgl2 mRNA and to knock down both maternal and zygotic Lgl2 protein expression. Embryos injected with Lgl2 MO-1 (Sonawane et al., 2005) showed dose-dependent defects, including a curved body axis phenotype at 48 hpf (Fig. 1J,M), similar to some zebrafish cilia mutants (Ferrante et al., 2009; Fogelgren et al., 2011; Kramer-Zucker et al., 2005), and hydrocephalus (Fig. 1J, arrow) that is associated with defects in ependymal cilia (Kramer-Zucker et al., 2005). In some cases, embryos showed more severe phenotypes, including a dramatically shortened body axis (Fig. 1L,M). Lgl2 MOs were co-injected with p53 MO to avoid off-target effects and embryos injected with p53 MO alone (MO control injection) appeared normal at 48 hpf (Fig. 1I,M). Injecting Lgl2 MO-2 (Hava et al., 2009) also resulted in dose-dependent curved body defects and hydrocephalus (Fig. 1K,M). Co-injecting sub-optimal levels of these two different MOs recapitulated these effects (supplementary material Fig. S1A) and injecting 150 pg of MO-resistant lgl2 mRNA partially rescued Lgl2 MO body axis defects, indicating the specificity of this phenotype (Fig. 1M). Early embryonic phenotypes observed in Lgl2 MO-injected embryos, but not seen in penner zygotic mutants, are consistent with MO interference with both maternal and zygotic lgl2 expression (Sonawane et al., 2005).

Next, development of ciliated organs was assessed in Lgl2 MO embryos by conducting immunofluorescence staining using aPKC antibodies to detect apical epithelial cell membranes and acetylated tubulin (aTubulin) antibodies to label cilia (Amack et al., 2007). In KV at the 8-somite stage, aPKC localized at apical membranes in Lgl2 MO embryos, indicating that apical-basal polarity was intact. However, the lumen of KV was smaller, and the number and length of KV cilia appeared to be reduced relative to controls (Fig. 1N,O). Similarly, at 24 hpf, otic vesicle lumen size was reduced in Lgl2 MO embryos compared with control embryos, and the number and length of motile cilia appeared to be reduced (Fig. 1P,Q). Also at 24 hpf, pronephric duct cilia appeared to be shorter and more disorganized in Lgl2 MO embryos, compared with the densely packed network of cilia in the lumen of controls (Fig. 1R,S). Together, these results indicate that Lgl2 is involved in the development of organs containing motile cilia.

To study the function of Lgl2 in more detail, we analyzed the development of KV, the first ciliated organ to appear in the zebrafish embryo. KV is analogous to ciliated structures in other vertebrates, including the ventral node in mouse (Nonaka et al., 1998), gastrocoel roof plate in frog (Schweickert et al., 2007) and notochordal plate in rabbit (Okada et al., 2005), that generate an asymmetric fluid flow involved in establishing left-right asymmetry in the embryo. KV is derived from dorsal forerunner cells that cluster near the end of gastrulation, become polarized epithelial KV cells and assemble one motile cilium per cell, which protrudes from the apical surface into a nascent lumen (Cooper and D’Amico, 1996; Essner et al., 2005; Kramer-Zucker et al., 2005). The lumen then expands during somitogenesis stages and the cilia beat to generate coordinated flow (Essner et al., 2005; Kramer-Zucker et al., 2005). To visualize KV, we used transgenic Tg(sox17:GFP) embryos, which express GFP in KV cells (Sakaguchi et al., 2006), and labeled cilia with acetylated tubulin antibodies (Fig. 2A,B). We found that the number (Fig. 2C) and length (Fig. 2D) of KV cilia were significantly reduced in Lgl2 MO embryos. In addition, the size of the KV lumen area measured at the middle focal plane of KV (the largest KV area) was smaller in Lgl2 MO embryos (Fig. 2E). Consistent with reduced KV lumen area in fixed Lgl2 MO embryos, measurements of the 3D volume of KV lumen in live embryos with GFP-labeled KV cells indicated that KV volume was significantly reduced in Lgl2 MO embryos relative to controls (supplementary material Fig. S2A). Despite the small lumen size, ZO-1 and F-actin, markers of apical membrane domains, were enriched at apical surfaces lining the lumen in Lgl2 MO embryos, similar to controls (supplementary material Fig. S2B,C), indicating that KV cells in Lgl2-depleted embryos establish apical-basal polarity. KV lumen and cilia defects in Lgl2 MO embryos were partially rescued by lgl2 mRNA (Fig. 2C-E), indicating that these effects were specific to loss of Lgl2 function.

The reduction in KV lumen size and number of KV cilia in Lgl2 MO embryos (Fig. 2B) suggested that there might be fewer cells in KV. Analysis of sox17 expression in precursor DFCs during epiboly showed that the size of the DFC domain was similar in Lgl2 MO embryos and controls. In a subset of Lgl2 MO embryos (23%, n=56), DFCs were not tightly clustered at 90% epiboly (supplementary material Fig. S3B-C), suggesting adhesion or migration defects that might affect the number of cells that incorporate into KV. However, counting GFP⁺ KV cells at the 8-somite stage in either Tg(sox17:GFP) embryos or a second transgenic line, Tg(dusp6:memGFP), that expresses plasma membrane localized GFP in KV cells (Wang et al., 2011a), revealed that the number of cells was similar in both control and Lgl2 MO embryos (Fig. 2F). Thus, Lgl2 knockdown disrupted lumen and cilia formation without affecting KV cell number.

Defective lumen formation in Lgl2 MO embryos suggested that the dynamics of KV morphogenesis might be disturbed. Therefore, time-lapse experiments were performed using Tg(dusp6:memGFP) embryos (data not shown) or double transgenic Tg(sox17:GFP; β;-actin:Lyn-TdTomato) embryos, in which cell membranes are labeled by Lyn-TdTomato (supplementary material Movie 1). An optical cross-section through the middle plane of KV was analyzed to visualize lumen formation between the 1- and 5-somite stages. In controls, GFP⁺ KV cells were dynamic and the lumen expanded during development (supplementary material Movie 1). Consistent with results in fixed embryos (Fig. 2B,E), Lgl2 knockdown disrupted lumenogenesis in live embryos (n=4/4 movies). KV cells formed a rosette-like structure in Lgl2 MO embryos, but the lumen failed to expand as it does in controls (supplementary material Movie 1). In addition, some KV cells in Lgl2 MO embryos appeared to detach from neighboring cells, suggesting defects in KV cell adhesion. Taken together, our analyses show Lgl2 regulates lumen formation and cilia formation in KV.

Consistent with disruption of KV development, Lgl2 MO embryos showed left-right patterning defects. Asymmetric looping of the heart, which normally looped to the right in controls, was often reversed or failed to loop in Lgl2 MO-1 and Lgl2 MO-2 embryos (Fig. 3A,B). These heart-looping defects were also observed in embryos co-injected with sub-optimal doses of the two different Lgl2 MOs (supplementary material Fig. S1B) and were partially rescued by lgl2 mRNA (Fig. 3B). Injecting lgl2 mRNA alone also disrupted heart looping (Fig. 3B), suggesting that normal left-right asymmetry requires tight regulation of Lgl2 activity. Examination of asymmetric expression of the Nodal-related gene southpaw (spaw), which is normally expressed in the left lateral plate mesoderm (Long et al., 2003), revealed dose-dependent abnormalities in Lgl2 MO embryos, including right-sided, bilateral and absent spaw expression (Fig. 3C,D). Analysis of additional asymmetrically expressed genes, pitx2c (Essner et al., 2000) and cyclops (Rebagliati et al., 1998) (also known as ndr2), and markers of the embryonic midline, sonic hedgehog (Krauss et al., 1993) and lefty1 (Bisgrove et al., 1999), indicated that Lgl2 depletion altered left-right patterning (supplementary material Table S1) without disrupting the midline that provides a barrier between left and right sides (supplementary material Fig. S4). These results uncover a crucial role for Lgl2 in left-right patterning of the embryo.

To assess where Lgl2 localizes in ciliated KV cells, antibodies specific for Lgl2 (Sonawane et al., 2009) were used in immunofluorescence staining experiments. In wild-type embryos, Lgl2 localized at the basolateral membrane of KV cells and was excluded from the apical domain marked by phalloidin staining of cortical F-actin (Fig. 4A). Lgl2 immunostaining was significantly reduced in KV cells of Lgl2 MO depleted embryos relative to controls (supplementary material Fig. S5), providing evidence for antibody specificity and MO efficacy. Interestingly, Lgl2 was also observed in some KV cilia, either throughout the cilium or at the cilium base (supplementary material Fig. S6A). To characterize this ciliary localization further, we analyzed motile ependymal cilia in the spinal cord and found Lgl2 localized to cell membranes and associated with a subset of these cilia (supplementary material Fig. S6B,C). As a second approach to study Lgl2 localization, the coding sequence for full-length Lgl2 was fused to enhanced yellow fluorescent protein (EYFP) cDNA to generate a fluorescent-tagged Lgl2:EYFP fusion protein. Similar to endogenous Lgl2, Lgl2:EYFP showed basolateral membrane localization in wild-type embryos (Fig. 4B). Expression of Lgl2:EYFP rescued axial phenotypes and heart-looping defects in Lgl2 MO embryos with similar efficiency as the wild-type Lgl2 (supplementary material Fig. S7), suggesting that the fusion protein is functional. Lgl2:EYFP was not detected in KV cilia, potentially owing to low abundance or access restrictions. However, Lgl2:EYFP was found to localize near the adherens junction protein E-cadherin (Fig. 4B), which is known to play important roles in KV development (Matsui et al., 2011; Oteíza et al., 2008; Oteiza et al., 2010).

Because the relationship between lumenogenesis and ciliogenic programs remain poorly understood, such that reduced ciliogenesis may be secondary to disrupted lumen formation, we focused on the role of Lgl2 in lumenogenesis. To begin to understand how Lgl2 regulates KV lumen formation, we analyzed E-cadherin. Lgl2 and E-cadherin have been previously shown to function antagonistically to control hemidesmosome formation in basal epidermal cells (Sonawane et al., 2009), and membrane localization of E-cadherin in the epidermis was disrupted in lgl2 mutant embryos (Reischauer et al., 2009). In addition, knockdown of E-cadherin in the DFC/KV cell lineage resulted in KV morphogenesis defects similar to those observed in Lgl2 MO embryos, including altered DFC clustering and small KV lumen size (Matsui et al., 2011; Oteíza et al., 2008; Oteiza et al., 2010). To study the relationship of Lgl2 with E-cadherin in KV, we performed immunostaining using E-cadherin antibodies and compared control and Lgl2 MO embryos. In control Tg(dusp6:memGFP) embryos, GFP labeled well-defined KV cell membranes and E-cadherin localized at lateral membrane domains (Fig. 4C). By contrast, KV cell membranes appeared disorganized in Lgl2 MO embryos and E-cadherin was reduced along the lateral domain (Fig. 4D). This indicates that Lgl2 is necessary for E-cadherin accumulation in KV membranes and that disruption of E-cadherin contributes to KV lumen defects in Lgl2 MO embryos.

To gain further mechanistic insight into how Lgl2 controls KV morphogenesis, we next tested whether Lgl2 is involved with vesicle trafficking machinery. Interestingly, the Rab family GTPase Rab11, which regulates recycling endosomes and vesicle trafficking from the trans-Golgi (Das and Guo, 2011), controls cilia length (Knödler et al., 2010) and the sorting and basolateral transport of E-cadherin in mammalian cell cultures (Lock and Stow, 2005). In zebrafish, MO knockdown of Rab11a has been recently shown to reduce KV lumen size (Westlake et al., 2011), similar to the KV phenotype observed in Lgl2 MO embryos. Thus, we hypothesized that Lgl2 genetically interacts with Rab11a to regulate KV development. To study Rab11a, we obtained the Rab11a MO and used an optimal dose to reduce KV size (supplementary material Fig. S8A-C) as previously reported (Westlake et al., 2011). Importantly, rescue experiments previously demonstrated this effect is specific to Rab11a depletion (Westlake et al., 2011). Quantification of KV lumen size revealed a significant difference between Rab11a MO embryos and controls (supplementary material Fig. S8D). In Rab11a MO embryos with small KVs, the cilia number and length were reduced (supplementary material Fig. S8E,F), suggesting that, similar to Lgl2, Rab11a is necessary for KV ciliogenesis. Consistent with these KV defects, cardiac left-right asymmetry was altered in Rab11a MO embryos (supplementary material Fig. S8G). Co-injecting MO-resistant mRNA encoding zebrafish Rab11a partially rescued left-right asymmetry defects in Rab11a morphants (supplementary material Fig. S8G), indicating that these phenotypes are specific to Rab11a knockdown.

The phenotypic similarities between Lgl2- and Rab11a-depleted embryos supported our hypothesis that Lg2 and Rab11a cooperate during KV morphogenesis, potentially to regulate vesicle trafficking of proteins essential for cilia formation and/or targeting E-cadherin to cell membranes. To test this hypothesis, genetic interaction experiments were conducted by co-injecting lower MO doses (Lgl2 MO-1^low + Rab11a MO^low) to determine whether partial knockdown of both Lgl2 and Rab11a resulted in synergistic effects on KV development. Although Lgl2 MO-1^low or Rab11a MO^low alone had modest effects on KV development, co-injection significantly reduced KV lumen volume (Fig. 5A,C). In contrast to lumen defects, the number and length of KV cilia were not affected in Lgl2 MO-1^low + Rab11a MO^low embryos (Fig. 5B,D,E). This indicates that KV lumenogenesis defects can be genetically uncoupled from cilia phenotypes.

We also investigated whether Rab11a genetically interacts with Lgl2 to regulate E-cadherin at lateral membranes of KV cells. Tg(dusp6:memGFP) embryos injected with Lgl2 MO-1^low or Rab11a MO^low showed accumulation of E-cadherin at KV cell junctions, as observed in control embryos (Fig. 6A-C). Quantification of E-cadherin staining along KV membranes in several embryos indicated that levels were modestly reduced in embryos injected with Lgl2 MO-1^low or Rab11a MO^low relative to controls (Fig. 6E). By contrast, co-injection of Lgl2 MO-1^low + Rab11a MO^low, which disrupted lumen formation, showed a more dramatic reduction of E-cadherin immunostaining at the lateral domain of KV cells (Fig. 6D) that was significantly different from embryos injected with Lgl2 MO-1^low or Rab11a MO^low alone (Fig. 6E). Taken together, our results show that genetic interactions between Lgl2 and Rab11a regulate E-cadherin and are crucial for normal KV lumenogenesis.

Motile cilia facilitate a number of crucial functions, including airway clearance, cerebrospinal fluid flow and the placement of organs along the left-right body axis. Here, we used KV in the zebrafish embryo to examine the role of Lgl2 in the development of ciliated organs. Lgl proteins are involved in several important biological processes, yet functions for Lgl during vertebrate development are unclear. We show that Lgl2 is involved in the genesis of motile cilia and cooperates with the Rab11a GTPase to regulate E-cadherin accumulation at cell-cell junctions and lumen formation. Importantly, reduction of both Lgl2 and Rab11a expression disrupted lumenogenesis without affecting ciliogenesis in KV, indicating that these processes can be uncoupled. Our study indicates that Lgl2 interacts with vesicular trafficking machinery to control lumen development in ciliated organs and that disrupting this pathway does not provide feedback that inhibits ciliogenesis.

The mechanism(s) by which Lgl proteins function at the molecular level has remained enigmatic, but several studies have used a loss-of-function approach or identified interactions between Lgl and specific protein complexes to shed light on Lgl functions. In this article, we present two novel roles for Lgl2 in ciliated organs: a Rab11a-associated role in lumenogenesis and a second role in ciliogenesis. In addition to lumen-expansion defects in Lgl2-depleted embryos, some KV cells in these embryos had short cilia whereas others were devoid of cilia. This observation indicates Lgl2 is in involved in formation of cilia (ciliogenesis) rather than (or in addition to) regulating cilia length. Cilia-associated phenotypes were not observed in Lgl1^-/- or Lgl2^-/- knockout mice, but Lgl1 and Lgl2 appear to be co-expressed throughout the early mouse embryo raising the possibility that Lgl1 and Lgl2 function redundantly to control cilia development in mammals. Ciliary defects are also apparently absent in zebrafish penner mutants, indicating that maternal Lgl2 compensates for loss of zygotic expression during early development. Our study provides the first evidence that Lgl2 plays early roles in the vertebrate embryo that are important for ciliated epithelium formation, body axis development and left-right patterning.

In Drosophila epithelial cells, Lgl proteins are known to interact with Scribbled and Discs large and antagonize the aPKC-Par3-Par6 complex to regulate apicobasal polarity. Loss of Lgl function in Drosophila or Lgl1^-/- knockout mice disrupts apicobasal polarity and results in hyperproliferation of specific cell types (Gateff, 1978; Klezovitch et al., 2004). Our analyses of apicobasal polarity in ciliated organs (Fig. 1; supplementary material Fig. S2) did not reveal gross defects in Lgl2-depleted embryos. Although malformations of the epithelia made it difficult to detect subtle changes in apical or basolateral membrane domains, the overall polarity appeared similar in Lgl2 MO and control embryos. In zebrafish penner mutants, loss of Lgl2 did not alter apicobasal polarity, but disrupted the formation of basally localized hemidesmosomes (Sonawane et al., 2005) and promoted epithlelial-to-mesenchymal transition (EMT) and over-proliferation of epidermal cells (Reischauer et al., 2009). In our studies of Lgl2-depleted embryos, we did not observe evidence of EMT or hyperproliferation of KV cells. Our results suggest that the functions of Lgl2 during ciliogenesis and lumenogenesis in developing epithelia are distinct from its roles in the zebrafish epidermis.

Lgl homologs interact with vesicle-trafficking components and exocytic machinery and are involved in delivering specific proteins to target membranes. Mutations in the yeast Lgl homologs Sro7 or Sro77, which interact with t-SNAREs, disrupt fusion of post-Golgi vesicles with plasma membrane (Gangar et al., 2005; Zhang et al., 2005). In zebrafish, Lgl2 acts synergistically with Clathrin-interactor 1 (Clint1), which binds clathrin-coated components and vesicle-associated SNARE proteins (v-SNAREs) involved in vesicle fusion (Dodd et al., 2009). We propose that Lgl2 mediates polarized vesicle trafficking, which is essential for development of ciliated organs (Fig. 7). This model includes trafficking of E-cadherin to basolateral membranes of KV cells. E-cadherin is highly expressed in the DFC/KV cell lineage (Kane et al., 2005; Oteiza et al., 2010) and is essential for normal KV development (Oteiza et al., 2010). As depletion of E-cadherin in DFC/KV cells disrupts DFC adhesion and KV lumen formation, loss of E-cadherin at basolateral membranes in Lgl2-depleted embryos is likely to be responsible, at least in part, for KV lumen defects. E-cadherin-based adherens junctions might be necessary for epithelial remodeling and/or to maintain epithelial integrity during fluid influx and lumen expansion.

In addition to a role in lumen formation, our functional analyses have uncovered a second role for Lgl2 in cilia formation (Fig. 7). We propose that Lgl2 regulates protein trafficking to cilia, which is necessary for normal ciliogenesis. Our detection of Lgl2 near some (but not all) cilia in fixed tissues (supplementary material Fig. S6) might reflect transient and dynamic localization of Lgl2 with cilia, which would be consistent with a role for Lgl2 in vesicle trafficking to cilia. Although we did not detect Lgl2:EYFP fusion protein associated with cilia in live embryos, generating stable transgenic lines that express Lgl2:EYFP coupled with higher resolution imaging may be a useful approach to study Lgl2 protein dynamics. In addition, it will be interesting to test our model that cilia formation can be separated from lumen formation by identifying new factors that regulate lumen size and then drawing correlations with ciliogenesis. It will also be interesting to test the role of Lgl2 in ciliogenesis in a cell culture system in the absence of lumenogenesis.

Similar to the roles of Lgl2 we have described in zebrafish, Rab11 mediates ciliogenesis and lumenogenesis in mammalian cells by regulating exocyst function (Bryant et al., 2010; Das and Guo, 2011; Datta et al., 2011; Knödler et al., 2010; Stenmark, 2009). The exocyst protein Sec15 (also known as Exoc6) has been shown to be a downstream effector of Rab11a, and Sec15 knockdown inhibited lumen formation in cultured Madin-Darby canine kidney (MDCK) epithelial cells (Bryant et al., 2010). In polarized MDCK cells, Rab11a is associated with recycling endosomes and is involved in targeting E-cadherin to cell junctions (Desclozeaux et al., 2008; Oztan et al., 2007). Our co-knockdown studies suggest that Lgl2 and Rab11a cooperate in the zebrafish embryo to localize E-cadherin to adherens junctions in KV cells. Thus, it is plausible that Lgl2 regulates the asymmetric distribution of membrane proteins such as E-cadherin via interactions with Rab11a, and potentially the exocyst complex, to regulate polarized KV tissue architecture and remodeling that is coupled with lumenogenesis. It is interesting that this process does not impact cilia formation in KV, indicating that cilia number and length is not influenced by lumen size in this organ. However, we cannot rule out an interaction between Lgl2 and Rab11 that is involved in ciliogenesis, owing to technical reasons (our MO doses may have not been sufficient to uncover such an interaction) or owing to other zebrafish Rab11 proteins, including Rab11al, Rab11ba and Rab11bb (Clark et al., 2011), which may compensate for loss of Rab11a during cilia formation.

In Drosophila, Lgl binds to myosin II and negatively regulates myosin activity (Strand et al., 1994), whereas mammalian Lgl2 positively regulates myosin II activity in MDCK cells (Wan et al., 2012). Thus, KV lumen expansion defects in the absence of Lgl2 might be partly due to aberrant actomyosin activity in KV cells. Regulation of actomyosin activity and E-cadherin accumulation may need to be coordinated (Shewan et al., 2005; Smutny et al., 2010) in order to control the KV cell-cell adhesions and contractility necessary for lumen formation. Additional experimental approaches will be needed to try to tease apart the roles of Myosin II in cell adhesion and contractility in ciliated organs and to test whether Lgl2 regulates either of these activities in the embryo.

In summary, this study has revealed new functions for Lgl2 that are important for the formation of ciliated organs and development of the embryo. Lgl2 controls formation and length of motile cilia, cooperates with the vesicle trafficking regulator Rab11a to mediate lumen formation and is necessary for proper left-right patterning and organ laterality. These results provide a framework for future work investigating the role of Lgl2 in vesicle trafficking during embryonic development.

We thank members of the Amack lab for helpful discussions and Mahendra Sonawane for donating reagents.