Kupffer's vesicle is a ciliated organ of asymmetry in the zebrafish embryo that initiates left-right development of the brain, heart and gut

The external body plan of vertebrates is bilaterally symmetric, but many internal organs, including the heart, digestive organs and parts of the brain,display highly conserved left-right (LR) orientations that are essential for their functions. Members of the nodal and leftycell-signaling families and pitx2, a bicoid-like transcription factor, display similar asymmetric expression patterns in the lateral plate mesoderm in chick, mice, frog and zebrafish embryos (reviewed by Yost, 1999). These genes are also asymmetrically expressed in the zebrafish brain during development(Bisgrove et al., 1999; Essner et al., 2000; Rebagliati et al., 1998; Thisse and Thisse, 1999). Perturbations of these asymmetric gene expression patterns correlate with alterations in brain, heart and gut LR morphogenesis.

A central question in developmental biology is what mechanism initiates asymmetric gene expression and subsequent asymmetric organogenesis? In mice,an elegant model has been proposed in which monocilia protruding from cells in the late gastrula node direct an asymmetric flow of extracellular fluid that results in the establishment of asymmetric gene expression(Nonaka et al., 1998). This`nodal flow' model is supported by the observations that LR defects are caused by mutations in microtubule motor proteins that affect either ciliogenesis,such as Kif3a (Marszalek et al.,1999; Takeda et al.,1999), Kif3b (Nonaka et al.,1998) and Polaris (Murcia et al., 2000), or cilia motility, such as Left-right dynein (Lrd),which is encoded by the iv gene(Supp et al., 1999). The strongest arguments for a role for nodal flow in the establishment of LR asymmetries come from cultured mouse embryos in which externally applied rightward fluid flow can reverse LR development in wild-type embryos and externally applied leftward flow can rescue LR development in mutants that would normally have inverted LR orientation(Nonaka et al., 2002).

The nodal flow model was initially proposed to move a peptide-signaling factor, perhaps a member of the FGF, TGFβ or Hedgehog signaling families. More recently, nodal flow has been proposed to function by asymmetrically activating non-motile mechanosensory cilia on the periphery of the node and initiating an asymmetric Ca²⁺ flux(McGrath et al., 2003)(reviewed by Tabin and Vogan,2003; Yost, 2003). Nodal flow is the earliest known event in murine LR development. However,genes implicated in ciliated node cell function are also expressed in non-node cells, and mutations of these genes give rise to a variety of laterality phenotypes and more pleiotropic phenotypes, making it impossible to exclude other mechanisms for LR development during early embryogenesis (for reviews,see Tabin and Vogan, 2003; Wagner and Yost, 2000; Yost, 2003). Furthermore, it is not known whether nodal flow is a murine-specific mechanism. Cilia formation and expression of lrd homologues have recently been observed in structures analogous to the node in chick, frog (Xenopus laevis) and zebrafish embryos (Essner et al., 2002), but the existence of motile cilia and nodal flow has not been demonstrated in non-murine embryos. Further confounding the issue, there are molecular asymmetries that precede the appearance of lrd expression and monocilia in Xenopus(Kramer et al., 2002; Kramer and Yost, 2002; Levin et al., 2002) and perhaps in chick (Levin et al.,1995; Stern et al.,1995). This raises the issue of whether ciliated cells in other vertebrate embryos generate fluid flow and have a conserved function for ciliogenesis genes in LR development.

In zebrafish, ciliated cells arise in the tailbud at the end of gastrulation (Essner et al.,2002) in a transient spherical organ called Kupffer's vesicle(KV). KV, first described in 1868(Kupffer, 1868), is a conserved structure among teleost fishes. Electron microscopy studies in the bait fish Fundulus heteroclitus have shown that a single cilium (i.e. monocilium) protrudes from each cell lining KV into the lumen(Brummett and Dumont, 1978). In zebrafish, KV is formed from a group of approximately two-dozen cells, known as dorsal forerunner cells (DFCs), that migrate at the leading edge of the embryonic shield (the zebrafish equivalent of the mouse node) during gastrulation. In contrast to other cells in this region, DFCs do not involute during gastrulation, but remain at the leading edge of epibolic movements. At the end of gastrulation, DFCs migrate deep into the embryo and organize to form KV (Cooper and D'Amico,1996; D'Amico and Cooper,1997; Melby et al.,1996). During subsequent somite stages, KV is found ventral to the forming notochord in the tailbud and adjacent to the yolk cell. Although KV was first described well over 100 years ago, it remains unknown whether DFCs and KV are mesodermal or endodermal in origin, and it is unclear what role they play during development, leading to the categorization of KV as an embryonic `organ of ambiguity' (Warga and Stainier, 2002).

Recently, using a novel technique to knockdown gene expression specifically in DFCs, we reported the first evidence that DFCs and/or KV function in LR patterning (Amack and Yost,2004). Here, we show that cilia that arise inside KV are motile and generate a consistent counterclockwise fluid flow. A combination of laser ablations, embryological manipulations, analyses of mutants and antisense morpholinos against zebrafish left-right dynein-related1(lrdr1) injected either into all embryonic cells or specifically targeted to DFCs, demonstrates that ciliated KV cells control LR development of the brain, heart and gut. The presence of motile cilia in zebrafish embryos supports the idea that fluid flow is a conserved LR mechanism used in all vertebrates. In addition, we propose a multi-step genetic pathway in which both the T-box transcription factor no tail (ntl) and components of the Nodal signaling pathway are necessary, either directly or indirectly, for the expression of lrdr1 in DFCs and for the morphogenesis of KV. Based on these analyses, we propose that KV is a transient embryonic `organ of asymmetry' that regulates the earliest known step in LR axis specification in zebrafish. These results comprise the first evidence that ciliated cells function during LR development in a non-murine embryo.

Danio rerio were maintained at 28.5°C on a 14 hour light/10 hour dark cycle. Wild-type zebrafish stocks were obtained from Ekkwill Breeders(Gibsonton, FL). Stocks carrying sur^ty686, sur^ty768,cyc^tf219 and cas^ta56 mutations were originally produced at the Max-Plank-Institute Fur Entwicklungsbiologie; oep^m134 at MGH/Harvard; ntl^b195 at the University of Oregon; and flhⁿ¹ at Newcastle (UK). Embryos were collected from natural spawnings, cultured and staged as previously described(Westerfield, 1995).

DFCs were ablated from wild-type embryos between shield and 80% epiboly stages using a VSL-337ND-S Tunable Nitrogen Laser (Laser Science) set to 440 nm. The laser was interfaced to a Leica DMRA compound microscope through a fiber optic cable (Photonic Instruments). Heart laterality was scored in DFC-ablated embryos (and all other embryos in this study) by scoring heart looping as normal, reversed or midline in anesthetized embryos at 2 days post-fertilization.

Embryonic dissections were carried out on agarose pads using 29 gauge needles on dechorionated embryos. Control embryos were also dechorionated at the same time and kept under similar conditions. All dissections were carried out stereotactically in that both the left and right sides were disrupted. Embryos that showed any disruption of the yolk cell were discarded. Embryos with normal ntl staining in the notochord were selected for LR marker analysis.

To isolate the 5′ end of zebrafish lrdr1, also referred to as dynein axonemal heavy chain 11 (Unigene # Dr20984), we performed 5′ RACE (rapid amplification of cDNA ends). lrdr1 cDNA sequence was compared with a genomic DNA contig in the Sanger Institute genome database(www.ensembl.org)that contained the 5′ end of the zebrafish lrdr1 gene in order to assign exon and intron boundries for the first seven exons. An updated version of the zebrafish genome from Ensembl includes an annotated fragment of lrdr1 containing the first 10 exons (Gene ID: ENSDARG00000004221). Antisense morpholino oligonucleotides (MO) directed against the lrdr1start codon (lrdr1-AUG MO:5′-GCGGTTCCTGCTCCTCCATCGCGCC-3′), the lrdr1 exon 2/intron 2 splice site (lrdr1E2/I2 MO:5′-ACTCCAAGCCCTCACCTCTTATCTA-3′), the start codon of lefty1 (lft1 MO:5′-GGCGCGGACTGAAGTCATCTTTTCA-3′) and a standard negative control MO (5′-CCTCTTACCTCAGTTACAATTTATA-3′) were obtained from Gene-Tools. To knock down gene expression in all cells, MO were injected between the one- and four-cell stages as previously described(Nasevicius and Ekker, 2000). To deliver MO specifically to DFCs, fluorescein-labeled MO were injected into the yolk cell between the 512-cell and 1000-cell stages, and embryos with fluorescent DFCs were selected for analysis, as described(Amack and Yost, 2004). RT-PCR was used to determine the effect of lrdr1E2/I2 MO on lrdr1mRNA splicing. Total RNA was extracted from embryos using TRIzol Reagent(Invitrogen) and cDNA was synthesized by MMLV-RT primed by random decamers(Ambion Retroscript kit). lrdr1 cDNA was amplified by 35 cycles of PCR using a forward primer (5′-ACATTCACGCCCTTCAAAAC-3′) in exon 2 and a reverse primer (5′-ACGTCCTGGATCATTTTTGC-3′) in exon 4. Amplicons were confirmed by DNA sequencing.

Antisense RNA probes were transcribed in the presence of digoxigenin-11-UTP or fluorescein-12-UTP (BMB) from linearized DNA templates using the Maxiscript kit (Ambion). Free nucleotides were removed using Bio-Gel P-6 Micro Bio Spin columns (BioRad). cDNA templates used included ntl(Schulte-Merker et al., 1994), pitx2 (Essner et al.,2000), shh (Krauss et al., 1993), lft1 and lft2(Bisgrove et al., 1999). Different probes and combinations of probes were used for lrdr1: the previously reported lrdr1 RT-PCR product(Essner et al., 2002), a genomic HindIII to HindIII fragment subcloned from a PAC corresponding to a region 5′ to the RT-PCR product, and a 5′ RACE PCR product. All lrdr1 probes produced similar results individually. In situ hybridizations were performed as previously described(Essner et al., 2000), and in some cases were automated using a Biolane HTI in situ machine (Huller and Huttner AG). Embryos were cleared in 70% glycerol in PBST and photographed with a Leica MZ12 stereoscope using a Dage-MTI DC330 CCD camera.

Embryos were fixed overnight in 4% paraformaldehyde at 4°C and dehydrated stepwise into methanol for storage at -20°C. After stepwise re-hydration, embryos were blocked for 1 hour in 5% goat serum (Sigma), 2%bovine serum albumin (BSA, Sigma) and 1% DMSO in PBST. Embryos were incubated with an anti-acetylated tubulin antibody (1:400, Sigma) and anti-Ntl antibody(1:100, a gift from D. J. Grunwald)(Schulte-Merker et al., 1992)in blocking solution. After washing embryos in 2% BSA and 1% DMSO in PBST,embryos were incubated with FITC-labeled goat anti-mouse IgG2b antibodies(1:500, Southern Biotech) and Alexa 568-labeled goat anti-rabbit antibodies(1:500, Molecular Probes). After washing, embryos were cleared in Slow Fade(Molecular Probes) and the tail region was removed and mounted. Embryos were imaged using an Olympus Fluoview scanning laser confocal microscope with a 60× objective.

Live embryos were mounted in 2% methylcellulose in a depression slide and cilia were imaged using a Zeiss Axioskop microscope with a 63× DIC objective. Movies were captured with a Photometrics Coolsnap HQ digital camera and Metamorph imaging software.

Embryos were removed from their chorions and mounted in 1% low melt agarose in system water on coverslips. Fluorescent beads (0.5-2 μm) (Polysciences)were pressure injected into KV at the six-somite stage. Embryos were imaged on a Zeiss Axioskop 2 microscope with a 10× Plan Apo objective. Movies were taken with a Nikon Coolpix 995 digital camera and manipulated using QuickTime Pro and Photoshop 7.

In mouse, Lrd (Dnahc11 - Mouse Genome Informatics) is an axonemal dynein heavy chain gene that is expressed in ciliated ventral node cells and is required for ciliary function and the generation of nodal flow(Supp et al., 1999). In order to identify the embryonic domains that might have a conserved ciliary function during zebrafish LR development, the expression pattern of the left-right dynein-related1 (lrdr1) gene(Essner et al., 2002) was examined by RNA in situ hybridization at several embryonic stages prior to the onset of LR asymmetric gene expression(Fig. 1A-D). Maternal lrdr1 mRNA was ubiquitous (see Fig. S1 in the supplementary material)prior to the onset of zygotic transcription (512-cell stage), but by late gastrula stages (70-90% epiboly), expression was restricted to dorsal forerunner cells (DFCs) (Fig. 1A). Double in situ analysis of lrdr1 and lefty1(lft1), a marker of dorsal margin and midline cells, was used to confirm that lrdr1 is expressed exclusively in DFCs at 70-90% epiboly(Fig. 1B) and revealed that lft1 is not expressed in DFCs as previously reported(Bisgrove et al., 1999) but in cells adjacent to DFCs in the embryonic shield/organizer. Sections through embryos stained for either lrdr1(Fig. 1C) or lft1(Fig. 1D) confirmed these complementary expression patterns. Neither lrdr1 nor lft1was detected in yolk cells.

Consistent with the fate of the DFCs, lrdr1 expression was observed in KV (Fig. 1E-G)during early somite stages (SS). lrdr1 mRNA was also detected in the mesoderm of the tailbud at 8 SS (Fig. 1F), in the notochord and floorplate at the embryonic midline and intermediate mesoderm during the formation of the pronephric ducts at 11 SS(Fig. 1G). Similarly, murine lrd is expressed in both node and non-node cells in mouse embryos(Supp et al., 1999). The expression of lrdr1 in DFCs and KV suggests that these tissues could have an analogous role to the ventral mouse node during LR development.

The progeny of DFCs in KV assemble cilia similar to those observed at the mouse node (Essner et al.,2002). To examine the developmental timing of cilia formation in DFC/KV, we used anti-acetylated Tubulin antibodies to label cilia at different stages during gastrulation and somitogenesis. To aid in the identification of DFCs and KV in whole embryos, anti-Ntl antibodies were used to label nuclei in the notochord, mesoderm and DFCs during gastrulation and in the notochord,tailbud and KV during somitogenesis. Cilia were not observed in the DFC domain during gastrulation (data not shown) but were detected at 4 SS in Ntl-positive cells under the notochord, in the region of KV(Fig. 1H). These cilia initially appeared as short projections into the lumen of the forming KV. Cilia in the well-formed KV were elongated at 10 SS(Fig. 1I) and remained elongated at 15 SS (data not shown). At 18 SS, the lumen of the KV was smaller or absent, although some of the dispersed KV cells remained ciliated(Fig. 1J). Confocal analyses of embryos at 6-8 SS (n=5) revealed a 1:1 ratio between the number of cilia visualized with the anti-acetylated tubulin antibody and the number of nuclei staining positive for Ntl in KV, suggesting KV cilia are monocilia.

To determine whether KV cilia are motile, we used videomicroscopy to examine KV in living embryos. At 6-10 SS, cilia were motile in the lumen of KV(see Movie 1 in the supplementary material). To test whether these cilia are capable of generating fluid flow inside KV, fluorescent beads were injected into KV at the 6 SS. These beads revealed a consistent directional counterclockwise flow inside KV (see Movie 2 in the supplementary material). Flow persisted until 10-12 SS, when asymmetric expression of the Nodal-related gene southpaw begins in left lateral plate mesoderm cells at the anterioposterior level of KV (Long et al.,2003). These observations suggested that cilia in KV might function in a manner analogous to cilia in the mouse node during LR development.

Murine genes implicated in nodal flow are also expressed in non-ciliated cells, and mutations in these genes, which result in a diverse array of laterality phenotypes and pleiotropic defects in other tissues, do not provide direct evidence that ciliated cells have a role in LR patterning. To directly test whether ciliated cells are required for LR patterning, DFCs were ablated from wild-type zebrafish embryos during gastrulation using a nitrogen laser. To verify that the ablation technique was efficient, embryos were fixed following ablation at 60% epiboly, and stained for ntl expression(Fig. 2B) or lrdr1expression (data not shown). With this laser-ablation technique, each cell is individually ablated at a specified focal plane, so it is unlikely that underlying yolk nuclei or adjacent mesoderm cells are affected. Ablated embryos lacked ntl (n=8/8) and lrdr1(n=7/8) gene expression in the DFCs, but had wild-type expression of ntl in adjacent mesoderm cells(Fig. 2B). Control embryos displayed wild-type ntl expression in both the mesoderm and DFCs(n=5/5, Fig. 2A). These controls indicate that laser ablation provides an efficient approach to eliminate DFCs during gastrulation without altering gene expression in neighboring cells. To examine the effects of DFC ablation on KV formation,DFC-ablated and unablated control embryos were immunostained at 6 SS with anti-Ntl and anti-acetylated Tubulin antibodies(Fig. 2C,D). Although two out of six DFC-ablated embryos had a ciliated KV that appeared similar to control embryos (Fig. 2C), two out of six of the DFC-ablated embryos formed a dismorphic KV that was misshapen (data not shown) and two out of six of the embryos did not form KV(Fig. 2D). These results indicate that ablation can result in a range of effects on KV morphogenesis,which probably reflects the efficacy of DFC ablation.

As the embryonic midline has been shown to be essential for LR development in frogs (Danos and Yost,1995), zebrafish (Chen et al.,1997; Danos and Yost,1996) and mice (Meno et al.,1998; Yamamoto et al.,2003), we assessed whether DFC-ablated embryos had intact midline structures. Immediately after ablation, ntl expression was normal in adjacent prospective midline cells (Fig. 2B). At 6 SS, the notochord and floorplate were intact as detected by DIC microscopy (data not shown) and Ntl expression was normal in notochord nuclei (n=6/6, Fig. 2D). At 22 SS (Fig. 2E,F), the notochord and floorplate were intact and contiguous, as detected by DIC microscopy (data not shown), and ntl and lft1 expression in DFC-ablated embryos (n=16/17), similar to wild-type embryos (n=71/71). Another indication of intact midline is the specific expression patterns of downstream LR gene expression patterns. Disruption of notochord development results predominantly in bilateral expression of these LR markers in lateral plate mesoderm, as exemplified in ntl and flh mutants in zebrafish(Bisgrove et al., 1999; Bisgrove et al., 2000) and notochord surgical disruptions in Xenopus(Lohr et al., 1997). By contrast, DFC-ablated embryos, displayed relatively low rate of bilateral expression, concurring with the conclusion that midline formation is unperturbed by DFC ablation.

In contrast to normal anterioposterior and dorsoventral development,ablation of DFCs at 60-80% epiboly had significant effects on brain and heart LR development as assessed by the molecular markers lefty1(lft1) and lefty2 (lft2). lft1 and lft2 are normally expressed in the left dorsal diencephalon and left heart field, respectively (Fig. 2G,I). In the diencephalon, DFC-ablated embryos showed equal numbers of embryos with right-sided (35%, Fig. 2H) or absent lft1 expression (35%) and a few with bilateral or left-sided expression (Fig. 2K). Unablated controls showed wild-type expression of lft1 (89%, Fig. 2G,K). In the heart field,the majority (53%) of DFC-ablated embryos had inverted expression of lft2 (Fig. 2J,L). Taken together, our analyses revealed that 82% of DFC-ablated embryos had laterality defects (Fig. 2M),whereas only 11% of unablated control embryos had a laterality marker defect(n=71, data not shown). When heart loop orientation was scored, DFC ablation at 60-70% epiboly resulted in 24% reversed hearts (n=46) and DFC ablation at 70-80% epiboly resulted in 29% reversed hearts(n=14), with control, non-ablated siblings having 2-3% heart reversals (n=274).

The results from laser ablations of DFCs indicate that DFCs or KV cells are essential for normal LR patterning, but do not define the developmental stages at which this patterning occurs. To directly test the role of KV in the generation of laterality, KV was disrupted by microsurgery at distinct stages during somitogenesis. The efficacy of KV disruption was analyzed by immunostaining with anti-Ntl and Tubulin antibodies. In control embryos, cilia and KV cells were organized in a normal spherical pattern(Fig. 3C), whereas in KV disrupted embryos this morphology was severely perturbed(Fig. 3D,E, n=13/17). In embryos in which KV was disrupted between the 3-7 SS, notochord (assessed by ntl expression) and tail development was normal, but expression of the left-side markers pitx2 in the dorsal diencephalon and lateral plate and lft2 in heart primordia was altered(Fig. 3G,H,J). By contrast,when the KV disruptions were carried out at 13-15 SS, asymmetric expression of pitx2 and lft2 was normal(Fig. 3K). Consistent with the asymmetric gene expression data, heart laterality was reversed in 35% of embryos in which KV was disrupted at 3-7 SS (n=23), while KV disruptions at 10-15 SS resulted in a low rate of heart laterality disturbances (4%, n=45). Although these results do not exclude the possibility that DFCs play an earlier and separate role in LR development(i.e. during gastrulation) before giving rise to KV, they indicate that KV is a necessary organ for LR patterning at 3-7 SS but is dispensable after 10 SS.

Mutations in the lrd gene in mice result in indeterminate LR patterning. Although there is significant variation among the three mutant alleles (lrd^lgl, lrd^iv,lrd^ΔP1), lrd mutant homozygotes display a range of organ laterality defects, from normal situs in ∼50% of the mutants, to discordant, randomized organ orientations, to ∼40% showing a complete organ reversal (Brown et al.,1989; Supp et al.,1999; Supp et al.,1997). To assess the roles of lrdr1 in zebrafish,antisense morpholino oligonucleotides (MO) directed against the start codon(lrdr1-AUG MO) of the lrdr1 mRNA were injected into embryos to block translation and deplete Lrdr1 protein. Embryos injected with 2-4 ng of MO at the one- to four-cell stages developed severe axial defects and many died by 2 days postfertilization (dpf). Injecting a lower dose (1 ng) resulted in apparently normal gross development(Fig. 4B), including an intact midline, as assessed by expression of ntl in the notochord(n=66/73, Fig. 4E) and sonic hedgehog (shh) in the notochord and floor plate(n=26/27, data not shown) at 22-26 SS. In addition, KV morphogenesis was normal in MO injected embryos (n=78/92) similar to controls(n=88/100). Cilia in KV (Fig. 4H) also appeared normal in injected embryos (n=10/13) as compared with controls (n=14/16, Fig. 4G). However, lrdr1-AUG MO-injected embryos had an increased frequency of reversed heart looping (26%, n=61), indicating LR patterning was perturbed. Expression of lft2 in the heart field and lft1 in the diencephalon was also disrupted in these embryos(Fig. 4E, Table 1). lft1 often displayed a higher incidence of absence in both MO injected and control embryos, probably owing to variations in the timing of expression. The analysis of lft1 reveals a previously unrecognized role for Lrd in the regulation of brain asymmetry and taken together with the lft2analysis suggests that Lrd plays an evolutionarily conserved role in LR patterning.

The large size of lrd mRNAs (>14 kb) precludes conventional rescue experiments in which MO-induced phenotypes are corrected by co-injection of synthetic rescuing mRNA. As an alternative approach to test the specificity of lrdr1-AUG MO, we injected a distinct MO that targeted the splice donor site at the junction of exon 2 and intron 2(lrdr1E2/I2 MO). Splice site-blocking MO have been used successfully to disrupt splicing of specific transcripts in zebrafish embryos and,advantageously, the efficacy of splice site-blocking MOs can be measured by RT-PCR (Draper et al., 2001). Using primers in exons 2 and 4 of lrdr1, RT-PCR analysis of total RNA extracted from embryos at 2 dpf indicated that injecting lrdr1E2/I2 MO blocked the removal of intron 2 from lrdr1 transcripts and reduced the levels of correctly spliced lrdr1 mRNA in a dose-dependent manner(data not shown). The retention of intron 2 (confirmed by sequencing RT-PCR products, data not shown) introduces several stop codons in the lrdr1-coding sequence. Embryos injected with lrdr1E2/I2 MO had phenotypes similar to embryos injected with lrdr1-AUG MO. lrdr1E2/I2 MO (1.6 ng) injected at the one- to four-cell stages blocked splicing of lrdr1 mRNA in DFCs at 80% epiboly(Fig. 4J, lane 5), reversed heart looping in 27% of injected embryos (n=44) [when compared with 0% in uninjected embryos (n=31)], and disrupted expression of LR markers (Table 1). In lrdr1E2/I2 MO injected embryos, midline expression of ntl(n=22/23) and shh (n=45/46) was intact at 24-26 SS. Furthermore, KV appeared normal in live embryos (n=146/162) relative to controls (n=58/62) and anti-Tubulin antibodies detected KV cilia at 6-10 SS (n=5). In general, lrdr1E2/I2 MO had a weaker effect on LR development than did the translation-blocking lrdr1-AUG MO (Table 1). This could reflect a difference in the effectiveness of the two MOs or may be due to the presence of maternal lrdr1 transcripts, which are presumably already spliced before MO delivery and therefore not targeted by lrdr1E2/I2 MO. Either of these possibilities would be consistent with the low level of normally spliced lrdr1 mRNA that remains in lrdr1E2/I2 MO embryos (Fig. 6J, lane 5). Although lrdr1E2/I2 MO injections resulted in a less penetrant phenotype, the effects on LR patterning corroborate the lrdr1-AUG MO data and together show that lrdr1 is an essential gene for LR development in zebrafish.

In the mouse embryo, it is not known whether Lrd controls LR development through its expression in ciliated cells in the node or through its expression in non-node cells. To test the role of lrdr1 specifically in KV in zebrafish, we used a recently developed technique to knockdown targeted gene function specifically in DFCs and KV cells(Amack and Yost, 2004). MO injected into 1- to 4-cell embryos knocks down gene function in all embryonic cells (Nasevicius and Ekker,2000). By contrast, MO injected into the yolk cell at mid-blastula stages (between the 512-cell and 1000-cell stages) diffuses throughout the yolk and, in a subset of embryos, localizes in DFCs. This creates chimeric embryos in which the targeted gene function is knocked down in DFCs but not in other embryonic cells (Amack and Yost,2004). Fluorescein-tagged MO allowed fluorescence microscopy to be used to select embryos in which MO had loaded specifically into DFCs and KV. Injecting either 1 ng lrdr1-AUG MO or 1.6 ng lrdr1E2/I2 MO into the yolk at mid-blastula stages had a weak or no effect on LR development(Table 1). This may be due to inefficiencies of DFC-targeted injections, such as variability in the amount of MO that enters DFCs from the yolk and MO not entering all DFCs in a given embryo (see discussion), or to the presence of maternal mRNA or protein as discussed above. Consistent with these possibilities, RT-PCR analysis showed that although 1.6 ng of lrdr1E2/I2 MO injected at mid-blastula stages interfered with lrdr1 splicing in DFCs at 90% epiboly, this effect was reduced relative to injections at the one- to four-cell stages(Fig. 4J, lane 6 versus lane 5). To increase MO efficacy, 1 ng of lrdr1-AUG MO and 1.6 ng of lrdr1E2/I2 MO were co-injected into the yolk at mid-blastula stages. In four out of five experiments, co-injected embryos (referred to as DFC^lrdr1MO embryos) showed an increase in aberrant lft1 and lft2 LR marker expression(Table 1). In one experiment,however, 93% of DFC^lrdr1MO embryos had normal lft2 expression (data not shown), further highlighting the variability of DFC-targeting injections and/or MO efficacy. DFC^lrdr1MO embryos that showed LR defects appeared similar to control embryos at 2 dpf (Fig. 4C). Furthermore, DFC^lrdr1MO embryos had intact expression of shh (n=91/95) in the midline at the 24-26 SS, developed a normal KV (n=51/71) similar to controls(n=67/89) and formed KV cilia detected by anti-Tubulin antibody staining (Fig. 4I, n=4). These experiments suggest that lrdr1 plays a role in DFC/KV cells that is important for normal LR development, but do not exclude the possibility of additional roles for lrdr1 during LR patterning in other embryonic cells and/or the yolk cell.

Two controls were designed to test whether LR defects caused by injecting lrdr1 MO at mid-blastula stages were specific to lrdr1knockdown in DFCs and their progeny. First, we used MO against lft1,which is expressed in adjacent shield cells but not in DFCs during gastrulation (Fig. 1B-D). Consistent with a previous report (Feldman et al., 2002), 1.5 ng lft1 MO co-injected with 2 ng of fluorescein-tagged negative control MO at the one- to four-cell stages affected the midline and resulted in bilateral expression of lft1 in the brain (n=27/28) and lft2 in the heart field(n=28/28). Injecting lft1 MO + control MO into the yolk at mid-blastula stages resulted in predominantly normal lft1 and lft2 asymmetric expression (Table 1). This indicates that MO injected into the yolk at mid-blastula stages do not affect non-DFC cells (e.g. neighboring midline cells) and do not have non-specific effects on LR development. Second, to test whether lrdr1 MO must enter DFCs to affect LR development, we co-injected 1 ng of lrdr1-AUG MO and 1.6-2.5 ng of lrdr1E2/I2 MO into the yolk at late blastula stages (dome stage to 30% epiboly), when connections between the yolk and DFCs are thought to be closed(Cooper and D'Amico, 1996). In these embryos, MO diffused throughout the yolk but did not accumulate in DFCs,did not affect splicing of lrdr1 RNA, which is expressed only in DFCs at 90% epiboly (Fig. 4J, lane 7), and did not disrupt asymmetric expression of lft1 or lft2 (Table 1). Although this control does not exclude the possibility that yolk syncitial layer (YSL) nuclei underlying DFCs(D'Amico and Cooper, 2001; Kimmel and Law, 1985) might have a role in LR development, it indicates that if lrdr1 is expressed in yolk cells, it does not have a detectable role in LR development. Together, these results indicate that lrdr1 expression in DFCs, which proceed to make ciliated cells in the KV, is essential for normal LR development of the viscera and brain.

Loss of Lrd in mutant mice renders nodal cilia immotile(Supp et al., 1999). To determine if MO knockdown of zebrafish lrdr1 affected cilia motility and fluid flow in KV, fluorescent beads were injected into KV of lrdr1-AUG MO injected embryos. KV appeared normal in these embryos,but flow of fluorescent beads was reduced or absent(Table 2 and see Movie 3 in the supplementary material). Embryos injected with negative control MO showed the characteristic counterclockwise flow seen in wild-type embryos(Table 2). Importantly, lrdr1-AUG MO injected embryos from these experiments that were not injected with fluorescent beads showed a high rate (24%, n=58) of heart laterality reversal, whereas all embryos injected with negative control MO (n=61) had normal heart laterality. This indicates that, as in mouse, lrdr1 is required for cilia motility and provides evidence that fluid flow inside KV is required for zebrafish LR determination.

DFCs express lrdr1 and give rise to a ciliated KV that we have shown is essential for the earliest known step in zebrafish LR development. In order to discover genetic pathways that are upstream of this first step, we screened a collection of laterality mutants in zebrafish by five criteria: (1)formation of DFCs; (2) endocytic activity in DFCs; (3) ntl, squint(sqt) and lrdr1 gene expression in DFCs; (4) organogenesis of KV; and (5) ciliogenesis in KV. As in mouse mutants (reviewed by Wagner and Yost, 2000), the laterality phenotypes of these zebrafish mutants are different, suggesting either pleiotropic phenotypes or disruption of distinct steps in LR patterning. Although the relatively late developmental effects on asymmetric lateral plate markers and organ orientation have been described(Bisgrove et al., 2000; Chen et al., 1997; Chin et al., 2000; Concha et al., 2000; Danos and Yost, 1996; Gamse et al., 2003; Long et al., 2003), the embryonic site of action and primary cause of the laterality defects in each of these mutants remain unknown.

Zebrafish laterality mutants have been classified based on the disruption of the normally left-sided markers lft1, lft2 and pitx2(Bisgrove et al., 2000; Bisgrove et al., 2003). In the present study, we examined the following laterality mutants: class I mutants ntl and floating head (flh) have defects in notochord formation and bilateral expression of asymmetry markers; class II mutants spadetail (spt) and casanova (cas)have randomized and discordant expression of laterality markers; class III mutant cyclops (cyc) lacks asymmetric expression of lft1 in the diencephalon but has normal expression of other asymmetry markers in the heart and gut. Class IV mutants one-eyed pinhead(oep) and schmalspur (sur) fail to asymmetrically express lft1, lft2 or pitx2 in the diencephalon, heart and gut, probably owing to a direct disruption of the Nodal signaling pathway in responding cells (Gritsman et al.,1999; Pogoda et al.,2000).

DFCs were observed by DIC microscopy from 60% to 80% epiboly, and usually comprised 15 to 27 cells in wild-type embryos. DFCs have strong endocytic activity from the dome stage to 70% epiboly that can be assessed by staining with the fluorescent vital dye SYTO-11(Cooper and D'Amico, 1996). Each of the laterality mutants, with the exception of cas, had DFCs that were endocytically active and had normal ntl expression(Table 3). In our study, cas mutants had DFCs, but they were fewer (average 10.5 cells/embryo), flatter, more disperse and did not take up SYTO-11. This somewhat concurs with a previous report(Alexander et al., 1999) that cas mutants lack DFCs, and indicates that cas, a Sox-related transcription factor, is required for DFC formation and endocytic activity. With the exception of cas, the formation of DFCs, their organization and their endocytic activity during gastrulation do not appear to be correlated with later laterality defects in the mutants examined.

To identify genetic pathway(s) upstream of lrdr1, laterality mutants were examined for lrdr1 expression in DFCs during gastrulation. Strikingly, ntl mutant embryos, which lack the Ntl T-box transcription factor (homologous to mouse brachyury), either failed to express lrdr1 in DFCs or had only a few DFCs that weakly expressed lrdr1 (Fig. 5B and Table 3). DFCs in ntlmutants appeared normal, expressed the Nodal family member sqt and ntl RNA, and were able to uptake SYTO-11(Table 3). Another T-box transcription factor gene, spt, is expressed in DFCs (data not shown)but is not required for lrdr1 expression, as spt mutants display normal lrdr1 expression patterns(Fig. 5C and Table 3). cas mutants also lacked lrdr1 expression (Fig. 5D and Table 3),probably owing to aberrant DFC formation. These results indicate that Ntl, but not the closely related T-box transcription factor Spt, is upstream of lrdr1. This is the first identification of a transcription factor upstream of dynein expression in ciliated node-like cells in any vertebrate embryo.

Analyses of zebrafish cell-signaling mutants revealed that the Nodal signaling pathway is also upstream of lrdr1 expression. DFCs in zygotic oep mutants were properly formed, endocytically active and expressed ntl, but failed to express lrdr1(Fig. 5E and Table 3). Oep is an EGF-CFC co-receptor that is essential for the reception of Nodal signaling(Gritsman et al., 1999). Consistent with these results, lrdr1 mRNA was not detected in maternal-zygotic sur (mzsur) mutants (n=0/25) that lack a forkhead transcription factor (Foxa1) thought to directly respond to Nodal signaling (Pogoda et al.,2000). Zygotic cyc^tf219 mutants had normal ntl expression (n=56/56 embryos from heterozygote crosses)and lrdr1 expression (n=263/264). Cyclops is a distinct Nodal-related ligand (Rebagliati et al.,1998; Sampath et al.,1998). Together, this mutant analysis indicates that Nodal signaling through oep and sur, but not the zygotically expressed Cyclops member of the Nodal family, is upstream of lrdr1expression. As discussed below, these results identify two genetic pathways that are required, either directly or indirectly, for lrdr1 expression in the DFCs: the ntl pathway and the Nodal signaling pathway.

Last, we screened laterality mutants for KV organogenesis and ciliogenesis within KV. ntl mutants (identified by a lack of Ntl immunostaining)formed patches of cells with anti-Tubulin labeled cilia, but these embryos failed to develop a normal KV (Fig. 5G; Table 3). In spt mutants, lrdr1 expression was normal and tubulin immunostaining detected cilia, but KV was disorganized(Fig. 5H and Table 3). cas and oep mutants showed an absence of both cilia and KV(Fig. 5I,J; Table 3). Notably, ntl,spt, oep and cas mutants all lead to the same endpoint of defective KV morphogenesis, but block distinct steps leading to a functional KV and affect downstream LR gene expression in dramatically different manners.

Cilia located on the ventral surface of the mouse node are thought to initiate LR development by generating a nodal flow that establishes asymmetric gene expression. Previous work has suggested that a similar ciliary-based mechanism within the embryonic node/organizer/shield in chick, Xenopus and zebrafish embryos may function to generate LR asymmetry in other vertebrate species, but no direct functional analysis has been reported. In zebrafish, dorsal forerunner cells (DFCs) and their progeny in Kupffer's vesicle (KV) express lrdr1, a dynein gene family member,and KV cells are ciliated (Essner et al.,2002), similar to KV cells in another teleost fish, Fundulus heteroclitus (Brummett and Dumont,1978). Another component of cilia, Inversin, has been found in several vertebrates and has been implicated in LR heart orientation in zebrafish (Morgan et al.,2002b; Otto et al.,2003). Furthermore, pkd2, which encodes a component of a cation channel (Hanaoka et al.,2000) that localizes to cilia in mouse cells(Yoder et al., 2002), has been shown to function during zebrafish kidney development(Sun et al., 2004) and LR determination (data not shown). However, four observations leave open the possibility that nodal cilia are not the loci of activity for LR development in any vertebrate. First, genes implicated in cilia formation or function and LR development are expressed ubiquitously in early development, for example, lrd (iv) in mice (Supp et al., 1999), lrdr1 in zebrafish (this report) and inversin (Invs) in mice(Mochizuki et al., 1998; Morgan et al., 1998) and zebrafish (data not shown). Genes implicated in mouse nodal cilia function have not been knocked out exclusively in the mouse node, and it is clear from the pleiotropic mutant phenotypes that some of these genes have other roles in development that, when perturbed, could lead to defects in LR development(reviewed by Wagner and Yost,2000). Second, Inversin protein is found not only in cilia but in other intracellular locations (Eley et al., 2004; Morgan et al.,2002a; Nurnberger et al.,2002; Nurnberger et al.,2004), suggesting it has non-ciliary functions. Third, cilia are found in other embryonic tissues such as the embryonic midline(Sulik et al., 1994), which has been implicated in LR development in frogs, zebrafish and mice(Danos and Yost, 1995; Danos and Yost, 1996; Meno et al., 1998), suggesting functions ascribed to cilia genes in LR development might occur in non-node cells. Fourth, motor molecules have been implicated in earlier steps in LR patterning in Xenopus (Levin,2003).

Here, using zebrafish embryos, we provide the first evidence that ciliated cells function during LR development in a non-murine embryo. Laser ablation of DFCs, which perturbs KV morphogenesis, and surgical disruption of KV directly show that DFC/KV cells play an essential role in zebrafish LR patterning. We also show that cilia in KV are motile and generate a directional fluid flow that is severely impaired by MO knockdown of lrdr1, which consequently alters LR development. Finally, analysis of LR mutations in zebrafish has revealed novel roles for both the ntl and Nodal pathways in the regulation, either directly or indirectly, of lrdr1 expression for the regulation of LR development.

Using a recently developed technique that targets MO to DFC/KV cells(Amack and Yost, 2004), we have found that lrdr1 expression in DFC/KV cells is required for normal LR development. MO injected into the yolk cell at mid-blastula stages often enter DFCs, and not other cells, probably via cytoplasmic bridges first characterized by Cooper and D'Amico(1996) by loading fluorescent dextran into the yolk between the 1000-cell to oblong stages. Dextran accumulated in at least some DFCs in ∼60% of these embryos and was also often found in some non-DFC marginal cells, indicating that both DFCs and other margin cells are potentially in cytoplasmic confluence with the yolk. We have observed similar results injecting fluorescent-tagged MO into the yolk cell at mid-blastula stages, including a high degree of embryo-to-embryo variability. Fluorescence microscopy was used to select embryos in which MO have loaded into DFCs but not other embryonic cells. It is possible that small amounts of MO, below detection by fluorescent microscopy, accumulate in cells other than DFCs. To test this, we targeted DFCs with MO against another gene that is required for normal LR development, lft1, which is not expressed DFCs, but in cells neighboring DFCs(Fig. 1B-D). Eighty-two percent of DFC^lft1MO embryos showed normal asymmetric lft2 expression (Table 1), indicating MO injected into the yolk at mid-blastula stages do not affect non-DFC cells. In a separate control for DFC^MOexperiments, embryos injected with lrdr1 MO into the underlying yolk syncitial layer (YSL) cells at stages when bridges between the YSL and DFCs have closed (dome stage to 30% epiboly) have normal LR development. Although this does not exclude a possible role for the YSL in LR development, it indicates that lrdr1 (this study) and ntl(Amack and Yost, 2004) do not function in YSL during LR patterning. Taken together, the results from three approaches - DFC ablation, KV surgical disruption and lrdr1 knockdown in DFC/KV cells - demonstrate that the ciliated KV is an embryonic organ of asymmetry that is required for the earliest known step in the generation of LR asymmetry in zebrafish.

Our analysis of KV formation and function in zebrafish mutants provides a template for reconsidering distinct phenotypes seen in mouse and human laterality syndromes (Bisgrove et al.,2003). Based on our results, we propose five distinct steps in KV that precede the earliest known asymmetric gene expression patterns in zebrafish (Fig. 6). In the first step, DFCs are induced during the blastula period. As previously reported (Alexander et al.,1999) and shown here, cas mutant embryos, which are deficient for a novel Sox-related gene and lack all endoderm, also lack normal DFCs. The few DFCs present in cas mutants are misshapen and defective in: endocytosis; lrdr1, ntl and sqt expression;ciliogenesis; and formation of KV. As cas was originally isolated as a cDNA responsive to Nodal signaling(Dickmeis et al., 2001) and two Nodal-related ligands, sqt and cyc, are expressed in close proximity to DFCs (Rebagliati et al., 1998), the Nodal pathway is strongly implicated in the induction of the DFCs. However, as cyc^tf219 mutants form DFCs that are SYTO-11 positive and express lrdr1, it is likely that other members of the Nodal family, perhaps maternally stored, are important for the formation of DFCs.

The second step in the KV pathway is the activation of lrdr1expression in DFCs at the end of gastrulation. DFCs in mutants that lack oep, a Nodal signaling co-receptor, or mzsur, a foxa1 transcription factor in the Nodal response pathway, fail to express lrdr1. In addition to Nodal signaling, ntl mutant embryos fail to express lrdr1 in the DFCs, suggesting that Ntl transcription factor is upstream of lrdr1. As ntl mutants have a normal number of DFCs that stain with SYTO-11, express ntlmutant RNA and sqt RNA, and form cilia(Table 3), it is unlikely that the absence of lrdr1 expression in ntl mutants is due to non-specific changes in DFC determination. In some cases, ntl can be induced by the Activin/Nodal pathway(Bisgrove et al., 1999; Toyama et al., 1995); one could propose that Nodal signaling is upstream of ntl, which is then upstream of lrdr1. However, Ntl protein is expressed in oepmutants that lack lrdr1 expression, although there might be a subtle and undetectable decrease in Ntl in oep mutants. In addition, DFCs in ntl and oep mutants differ in sqt expression and cilia formation. Together, this suggests that the ntl-dependent induction of lrdr1 is a pathway parallel to the Nodal pathway. Obviously, mutant analysis indicates that lrdr1 expression is dependent on ntl and components of the Nodal pathway, but does not distinguish whether this dependence is direct or indirect. Clearly, both of these genetic pathways have other functions in KV development beyond regulation of lrdr1 expression(Fig. 6), as lrdr1morphants do not display defects in KV organogenesis and cilia formation seen in ntl and oep mutants. Conversely, although knockdown of lrdr1 disrupts directional fluid flow in the KV, one cannot exclude other roles for lrdr1 within DFC/KV cells for LR development.

In the third step, DFCs organize to form KV at the completion of gastrulation. This process involves ingression into the forming tailbud and formation of an epithelial sheet at the base of the notochord. KV formation is dependent on Nodal signaling components (exemplified by oep) and Ntl,as well as on additional genes that are not required for lrdr1expression, including the T-box transcription factor Spt. Using the DFC^MO technique, KV organogenesis was found to be cell-autonomously dependent on ntl (Amack and Yost,2004). It is not yet known whether defects in KV development described here (Table 3) are cell autonomous for spt and components of the Nodal pathway.

In the fourth step, cilia form on KV cells and project into the lumen. Surprisingly, cilia formation can be uncoupled from the normally coincident formation of KV in step 3, as spt and ntl mutants do not form an organized KV but form cilia (Fig. 5G,H; Table 3). In agreement with observations in Dnahc11/Dnahc11 mice(Supp et al., 1999), cilia formation in zebrafish does not appear to be dependent on lrdr1, as cilia are found in ntl mutants that lack normal lrdr1expression (Fig. 5G) and in lrdr1 morphant embryos (Fig. 4H). In the fifth step, cilia motility and fluid flow in KV is dependent on lrdr1, which determines normal asymmetric gene expression patterns in lateral tissues.

This mutant analysis points out that perturbation of one or combinations of the five steps in KV function outlined above results in altered LR development in the brain, heart and gut. Of the characteristics we examined, defects in KV organogenesis alone (e.g. spt), KV organogenesis plus lrdr1expression (e.g. ntl), KV ciliogenesis, organogenesis and lrd1 expression (e.g. oep), or all of the above plus DFC formation (e.g. cas) results in LR abnormalities. Interestingly, each of the mutants has distinct expression patterns of downstream laterality markers (Bisgrove et al.,2000). Although it might be tempting to speculate that defects in the distinct KV steps outlined above result in distinct laterality phenotypes in the rest of the embryo, it is important to note that the genes involved are expressed in other cells besides DFC/KV cells, and such a conclusion can not be made until the gene of interest is knocked down specifically in DFC/KV cells. For example, DFC^MO knockdown of ntl has a laterality phenotype that is distinct from bilateral expression patterns seen in ntl mutants and whole-embryo ntl morphants, indicating that ntl has distinct roles in DFC/KV and in midline cells for LR development (Amack and Yost,2004). Our surgical disruptions of KV indicate that KV is no longer needed for LR development by 10 SS, but do not exclude the possibility that different LR signals come from the DFCs or the KV at earlier stages in DFC/KV development.

Several models have been proposed to explain how cilia movement and fluid flow influences LR development, and how this information is transmitted to lateral plate mesoderm to initiate asymmetric expression of Nodal, Lefty and Pitx2 (for review (Mercola,2003; Raya and Belmonte,2004)). Flow has been proposed to influence the localization of a signaling molecule (Okada et al.,1999) or asymmetric activation of mechanosensory cilia and intracellular Ca²⁺ (McGrath et al., 2003) in mice. In chick, nodal cilia are present(Essner et al., 2002). Although motility and nodal flow have not been reported, extracellular Ca²⁺ has been proposed to activate lateral asymmetric nodal expression via a Notch-dependent pathway(Raya et al., 2004). Here, we show in zebrafish that lrdr1-dependent asymmetric fluid flow in KV probably results in asymmetric gene expression that sweeps from the tail to the head of the developing embryo. However, the mechanisms by which counterclockwise flow in KV is translated to molecular asymmetries near KV are not known.

In zebrafish, charon, an inhibitor of nodal signaling, is expressed in the tailbud at 2-3 SS during the formation of KV and then symmetrically adjacent to KV until the 14 SS(Hashimoto et al., 2004). Southpaw (spaw), a member of the nodal family, is bilaterally expressed immediately adjacent to charon(Hashimoto et al., 2004; Long et al., 2003). Morpholino knockdown of charon results in bilateral expression of spawand other asymmetric markers in lateral plate mesoderm (cyc, leftyand pitx2). Conversely, knockdown of spaw results in absence of downstream asymmetric markers. It is not known how the symmetric expression patterns of charon and spaw control subsequent asymmetric expression of spaw and other downstream genes. However, the bilateral spaw and charon expression near KV during the period that we have shown KV is important in LR development suggests that information derived from the lrdr1-dependent asymmetric fluid flow might influence the balance or interactions of spaw and its antagonist charon in adjacent cells.

There is elegance in the frugality of using the same signaling pathway at multiple stages in LR development. The T-box transcription factor Ntl independently drives lrdr1 expression and KV morphogenesis(Fig. 6) and midline formation in mesoderm cells, both of which are necessary for distinct aspects of normal asymmetric gene expression in lateral plate(Amack and Yost, 2004). Similarly, the Nodal signaling pathway is required for the induction of lrdr1 gene expression in DFCs in order to drive motile cilia in Kupffer's vesicle, and ciliated KV is necessary for driving normal LR asymmetric expression of components of the Nodal signaling pathway in lateral mesoderm. This novel aspect of the Nodal pathway, upstream of lrdr1function, suggests that analyses that have implicated the Nodal pathway in LR development in mice, chick, frog and zebrafish should be re-investigated for proximal defects in the formation and function of ciliated cells.

Several lines of evidence presented here indicate that KV cilia in zebrafish are required in LR development and are analogous to cilia located on the ventral mouse node. However, it is not known whether the generation of a flow in KV is the first symmetry breaking event in zebrafish. Certainly, the appearance of motile cilia in the KV, and the developmental period during which DFC/KV manipulations can alter LR development, precedes the appearance of asymmetric gene expression of spaw, cyc, lft1 and lft2 in lateral tissues. As such, it is the earliest known event in zebrafish LR development. However, recent evidence from frog embryos suggests that asymmetries are generated long before the appearance of ciliated cells(Kramer et al., 2002; Kramer and Yost, 2002; Levin et al., 2002). It is not yet known whether these mechanisms are amphibian specific, and future work will be focused on investigating potential connections between early molecular asymmetries and nodal flow. In chick, lrdr expression and cilia are found at stage HH4 (Essner et al.,2002), but more detailed expression analysis, cilia motility and fluid flow have not been reported, making it difficult to know whether early molecular asymmetries (Levin et al.,1995; Stern et al.,1995) precede a role for cilia. The results presented here demonstrate the importance of cilia in Kupffer's vesicle, a transient embryonic organ of asymmetry, for the establishment of left-right asymmetry in the gut, heart and brain in zebrafish. These data provide the first direct functional evidence of a ciliary-based mechanism operating in a non-murine vertebrate embryo during LR patterning.

We thank M. Brueckner for sharing mouse LRD sequences during our cloning of lrdr1, and B. Bisgrove, W. Branford and M. L. Condic for comments on the manuscript. M. Yoshigi and T. Cooke (from A. G. Heinze) provided technical assistance in imaging cilia in live embryos. This work was supported in part by grants from NIHLB and the Huntsman Cancer Foundation to H.J.Y. and an NRSA fellowship to J.D.A.