Origin and shaping of the laterality organ in zebrafish

A key step during the formation of the vertebrate body plan is the establishment of the left-right (LR) axis, which results in the characteristic asymmetric anatomy and positioning of the internal organs. Studies in several organisms have demonstrated that handedness of LR axis determination depends on the activity of a conserved embryonic ciliated structure (referred to as the organ of laterality), and involves a mechanism by which cilia beating generates a directed extracellular fluid flow (Nodal flow) that breaks the initial chemical and/or mechanical equilibrium of the embryo(McGrath et al., 2003; Nonaka et al., 1998; Tanaka et al., 2005).

Although the conservation and function of laterality organs are well established in vertebrates (Okada et al.,2005; Schweickert et al.,2007), comparatively little is known about their embryonic origin and the morphogenetic mechanisms leading to their formation. To obtain insights into these events, we analysed the formation of Kupffer's vesicle(KV), the zebrafish organ of laterality(Essner et al., 2005; Kramer-Zucker et al., 2005). KV is a transient organ proposed to arise from non-involuting, highly endocytic marginal blastomeres (deep NEM cells) that are positioned below surface-enveloping layer cells (EVL-NEM cells) at late blastula stage. Fate map studies have suggested that at the onset of gastrulation, deep NEM cells become displaced from the blastoderm margin to form a distinct group of dorsal forerunner cells (DFCs) (Cooper and D'Amico, 1996; D'Amico and Cooper, 1997; Melby et al.,1996). DFCs then move to the vegetal pole in close contact with the overlying EVL margin and become transformed into an epithelial vesicle(KV) that later produces a cilia-based flow within its interior lumen(Amack et al., 2007; Essner et al., 2005; Kramer-Zucker et al., 2005). During late somitogenesis, the vesicle collapses and DFCs become incorporated into notochord, somites and tail mesenchyme(Cooper and D'Amico, 1996). Despite these remarkably detailed studies, crucial questions remain concerning the cellular and molecular mechanisms that lead to DFC formation and the subsequent transformation of this cell group into a functional organ.

In this study, we demonstrate that DFCs are formed by ingression of surface epithelial cells at the dorsal margin of the zebrafish gastrula, and that Nodal/TGFβ signals are both required and sufficient to induce this process. We further show that, after ingression, a subset of cells within the DFC cluster retains its original polarisation through persistent apical contact with the overlying EVL. These pre-polarised cells serve as `seeds'around which the remainder of DFCs assemble and epithelialise to form the mature KV. These findings point to a conserved mechanism of Nodal/TGFb-induced outer epithelium internalisation during laterality organ formation and demonstrate a novel mechanism by which external apicobasal polarity is propagated from the surface epithelium to an interior organ during vertebrate development.

The following zebrafish (Danio rerio) lines and alleles were used:wild-type TL, Tg(sox17:GFP)(Sakaguchi et al., 2006) and Tg(β-actin:HRAS-EGFP)(Cooper et al., 2005). Embryos were grown at 31°C unless otherwise indicated, manipulated in E3 zebrafish embryo medium or Danieau buffer, and staged according to morphology(Kimmel et al., 1995) and age[hours post-fertilisation (hpf)].

For mRNA synthesis, pCS2+ expression vectors containing the cDNAs for different constructs were linearised with NotI and transcribed by SP6 mRNA polymerase as described (Montero et al., 2003). For mosaic expression, one marginal blastomere was injected at the 64-cell stage. mRNA amounts used for each experiment are given in the figure legends.

Embryos were fixed in 2% or 4% paraformaldehyde (PFA) and stained as described (Koppen et al.,2006). Stained embryos were mounted on agarose-coated dishes in phosphate buffered saline/0.5% Triton X-100 medium or embedded in 1%low-melting-point agarose. The following primary antibodies and dilutions were used: mouse anti-ZO-1 (Invitrogen, 1:200), rabbit anti-aPKC-ζ (Santa Cruz Biotechnology, 1:200) and mouse anti-acetylated α-tubulin (Sigma,1:400). Samples were imaged on a Zeiss LSM META confocal microscope using an Achroplan 40×/0.8 W dipping objective or a Plan-Apochromat 40×/1.2 W objective.

Embryos from Tg(sox17:GFP) and Tg(β-actin:HRAS-EGFP) in crosses or crosses between both lines were manually dechorionated and mounted in 0.5% low-melting-point agarose just prior to imaging. Time-lapse, multiple focal plane (4D)microscopy was then performed at 22°C or 28°C on a BioRad Radiance 2000 multi-photon or a Zeiss LSM META confocal system. Movies were processed using Volocity (Improvision) and ImageJ(http://rsb.info.nih.gov/ij/)software.

Lumen volume analysis was performed by generating 3D renderings of anti-aPKC-ζ staining using Volocity software. KV cilia length measurements were performed on z-projections of dorsal confocal stacks of Tg(sox17:GFP) embryos stained with anti-acetylatedα-tubulin using ImageJ software. The number of DFCs during epiboly was determined using Imaris software (Bitplane), employing the spot recognition function to count cell nuclei of sox17:GFP-expressing DFC clusters.

To address the mechanism that leads to the formation of the DFCs, we performed time-lapse multi-photon confocal imaging of sphere-stage zebrafish embryos (4 hpf) expressing the membrane marker Tg(β-actin:HRAS-EGFP). At this stage, the blastoderm consists of two main tissues: an outer epithelial monolayer referred to as the enveloping layer (EVL), and the deep cells that make up the embryo proper. Both tissues then undergo a vegetal movement in order to enwrap the yolk cell in a process known as epiboly (Kimmel et al., 1995) (see Fig. 1M for illustration).

As expected, we found that at the sphere stage, DFCs were not yet discernible at the future dorsal side of the embryo. Instead, we observed dorsal surface epithelial (DSE) cells that were in direct contact with the yolk syncytial layer and exhibited filopodia-like protrusions at their leading edge (Fig. 1A,G,J,M). Strikingly, we found that during subsequent development, a subset of marginal and submarginal DSE cells internalise through a process of ingression beneath the surface epithelium (Fig. 1;see Movies 1 and 2 in the supplementary material; data not shown). Ingression occurred as trailing cells of the EVL moved over the DSE cells, displacing them beneath the surface. Once ingressed, DSE cells were clearly recognisable as DFCs by their position below the EVL margin and ahead of the deep cell margin. We will therefore refer to these newly ingressed cells as DFCs. These cells most likely correspond to the previously described deep NEM cells(Cooper and D'Amico, 1996; D'Amico and Cooper, 1997). However, whereas deep NEM cells were observed to lie within the dorsal blastoderm margin, we found newly ingressed DSE cells to be positioned ahead of the deep cell margin. This discrepancy is most likely due to the different labelling and imaging methods used to identify DFCs. Together, we conclude that DFCs form via the ingression of marginal and submarginal DSE cells beneath the surface epithelium.

Importantly, we observed no incidences of deep cell conversion into DFCs(n=5 embryos; Fig. 1G-I; see Movies 1 and 2 in the supplementary material; data not shown). Moreover, the analysis of surface epithelial morphogenesis in lateral or ventral regions of the embryo revealed only very sporadic cell ingression(1/115 cells in 10 embryos; see Movie 3 in the supplementary material),indicating that coordinated ingression behaviour is restricted to DSE cells.

DSE cell ingression was often immediately preceded by a cell division(Fig. 1G-I,M; see Movie 2 in the supplementary material). Typically, the division occurred in the plane of the epithelium, giving rise to two ingressing cells. However, in younger embryos (around the `high' stage, 3.3 hpf), we frequently observed delaminating divisions of DSE cells, which gave rise to a DSE cell and a deep cell. We observed that in a few instances, the newly formed DSE cell became a DFC, while the second daughter cell exclusively became a deep cell (see Fig. S1 and Movie 4 in the supplementary material). This finding is surprising, as previous analysis of cell division patterns of the zebrafish blastula had suggested that a DSE cell resulting from a delaminating division is committed to the EVL fate (Kane et al.,1992). Together, we conclude that the type of DSE cell division does not strictly correlate with EVL versus DFC fate.

DFC formation took place concurrently with the onset of embryonic epiboly,during which marginal deep cells, marginal EVL cells and DFCs all advance towards the vegetal pole (Cooper and D'Amico, 1996; Melby et al.,1996). We found that the conversion of DSE cells into DFCs consistently completed by 50% epiboly, immediately prior to the formation of the embryonic organizer (shield). As epiboly progressed, the DFC group became increasingly separated from the dorsal marginal deep cells and remained in direct contact with the advancing dorsal EVL margin(Fig. 1G-I,M)(D'Amico and Cooper, 1997). Furthermore, the number of DFCs increased between shield stage and 80% epiboly from 31±9 (n=16 embryos, mean±s.d.) to 50±11(n=9), as determined using a Tg(sox17:GFP) transgenic line(see below). This increase is likely to be due primarily to cell division within the cell cluster, based on our analysis of time-lapse movies (data not shown). These numbers are higher than those determined previously(Cooper and D'Amico, 1996),probably because of the different methodologies applied to identify DFCs (see Materials and methods).

Ongoing with vegetal DFC progression, DFCs rapidly converged to the dorsal side by intercalating between each other, thereby transforming the initially wide array of cells into a more compact and oval-shaped cluster by about 80%epiboly (Fig. 1D-I; see Movie 5 and Figs S2 and S3 in the supplementary material). This convergence was notably mirrored by dorsal convergence of the overlying dorsal marginal EVL cells (Fig. 1A-C,G-I). Strikingly, a subset of DFCs became increasingly bottle-shaped during epiboly and remained persistently linked to the dorsal EVL margin via attachment points that were enriched for the tight-junction components ZO-1 (Tjp1 - ZFIN)and protein kinase C zeta (aPKC-ζ; Prkcz - ZFIN)(Fig. 2A-B′ and data not shown). By contrast, the remainder of cells in the cluster appeared unpolarised, forming dynamic cell protrusions without a preferential orientation (see Movie 5 and Fig. S3 in the supplementary material). This suggests that tight-junction based DFC-EVL attachment couples the vegetal movement and convergence of both tissues. This finding is consistent with previous analyses suggesting that the movement of DFCs and of the EVL are coupled via physical tissue linkage(D'Amico and Cooper, 1997; Solnica-Krezel et al.,1996).

To characterise the cellular events that transform the relatively unstructured DFC cluster into a highly organised vesicle after completion of embryonic epiboly, we made use of a transgenic fish line that strongly expresses GFP in DFCs via the sox17 promoter(Sakaguchi et al., 2006). Consistent with the above results, we found that multiple subgroups of bottle-shaped DFCs were attached to overlying dorsal marginal EVL cells via ZO-1-rich subcellular structures at 80% epiboly(Fig. 2C,C′). After reaching the vegetal pole during the final phase of epiboly, the cluster became separated from the overlying dorsal marginal EVL cells, most likely owing to dorsal marginal deep cells moving between the two tissues during the vegetal sealing of the embryo (Fig. 2D,D′; see Fig. S2 in the supplementary material). Using both live imaging and immunolabelling, we found that during this separation the apices of bottle-shaped DFCs turned towards the inside of the cluster and generated multiple three-dimensional rosette structures around ZO-1-rich apical focal points (Fig. 2D,D′,F,F′; see Movie 6 in the supplementary material). At the same time, the DFC cluster became increasingly compact and acquired the shape of a flattened sphere(Fig. 2C-D′; see Figs S2 and S3 in the supplementary material).

The separation of the DFC cluster from dorsal marginal EVL cells and its rearrangement into rosettes was quickly followed by the formation of an interior vesicle lumen (Fig. 3A-C; see Movie 7 in the supplementary material). This occurred as the initially widely spaced apical focal points within the cluster coalesced into a dense network around which cells arranged into a single large rosette structure, incorporating previously unpolarised DFCs. During these rearrangements, small extracellular lumina typically formed at multiple apical focal points (Fig. 3A,B,D,E,G,H) and rapidly fused into a single ZO-1-lined lumen(Fig. 3C,F,I). Lumen volume then increased approximately 12-fold between the 1-somite and 4-somite stages(see Fig. S4 in the supplementary material). Live imaging of DFCs expressing GFP revealed that cytoplasmic, vacuole-like structures fused with the apical membrane domain as the lumen formed, suggesting that apical vacuole fusion contributes to apical membrane expansion and lumen growth (see Movie 8 in the supplementary material). Interestingly, we found that lumen formation was directly coupled to ciliogenesis. Small, tubulin-rich cilia were first detected at the onset of lumen initiation, when the lumen-facing membrane domain of polarised DFCs started developing cilia(Fig. 3J-L). Lumen volume and cilia length continuously increased during KV maturation (between bud and 4-somite stage), while cilia number remained essentially constant (see Fig. S4 in the supplementary material). This suggests that no additional cilia-forming cells are recruited to the cluster during lumen formation and expansion. Together, these observations suggest that initial DFC polarisation is established during epiboly and contributes to the subsequent formation of rosette structures by a subset of polarised DFCs. Later epithelial differentiation to form KV involves concurrent lumen formation and ciliogenesis.

Our dynamic analysis of DFC formation revealed a principal mechanism that resembles the formation of the internal germ layers, which are formed through cell internalisation beneath the future ectoderm. Numerous studies have demonstrated that the Nodal signalling pathway acts as a key regulator of this process (reviewed by Schier and Talbot,2005). Importantly, recent reports have shown that DFC number increases or decreases in response to enhanced or reduced Nodal signalling,respectively (Choi et al.,2007). In agreement with this, we observed that overexpression of the Nodal antagonist Lefty strongly decreased the DFC number(Fig. 4B,E,H). In addition,maternal-zygotic one-eyed pinhead mutants, which are defective for Nodal signalling, completely lacked recognisable DFCs at 80% epiboly (data not shown). Together, this strongly suggests that Nodal signalling is required for surface cell internalisation to generate DFCs.

To test whether Nodal signalling acts instructively during DFC formation,we performed gain-of-function experiments by overexpressing the Nodal signalling ligand Cyclops (Cyc; Ndr2 - ZFIN). Strikingly, we observed that numerous DFC clusters were generated from DSE cells along the whole circumference of the embryo (Fig. 4C). A major dorsal cluster was typically observed that showed increased DFC number at 80% epiboly (104±50, n=5 embryos,mean±s.d.) compared with uninjected controls (50±11, n=9). The enlarged DFC clusters were attached to the EVL margin via apical cell domains, similar to in controls(Fig. 4A,C). Multiple DFC clusters were also observed at the bud stage and early somite stages. These clusters typically contained the characteristic interior apical focal points of controls and underwent partial lumen formation(Fig. 4F,I, and data not shown). These results indicate that enhanced Nodal signalling is sufficient to induce ectopic DFC clusters that undergo partially normal morphogenetic transformation.

To gain insight into the dynamics of ectopic DFC formation in Nodal-overexpressing embryos, we performed time-lapse confocal imaging of Tg(β-actin:HRAS-EGFP) embryos co-injected with cyc mRNA and fluorescently labelled dextran (to identify cyc-expressing cells). Importantly, we found that in these embryos,DFCs formed via the same principal mechanism observed in controls (see Fig. 1). Namely, surface epithelial cells located at the Cyc-overexpressing side of the embryo became transformed into DFCs-like cells through their displacement beneath the EVL sheet after sphere stage (Fig. 4J-L′; see Movie 9 in the supplementary material). In addition, we found this behaviour predominantly in the dextran-negative cells(n=3 movies), suggesting that Cyc induces cell ingression in a non-cell-autonomous manner. In summary, our findings indicate that DFC formation through ingression of surface epithelial cells is controlled by Nodal signalling, which plays both instructive and permissive roles in this process.

In this study, we have investigated the mechanisms leading to the formation of DFCs in the zebrafish embryo, and the subsequent transformation of this group of cells into KV. Our findings reveal that DFCs are formed through the ingression movement of a subset of surface epithelial cells on the dorsal side of the embryo (DSE cells). Ingression is initiated after sphere stage (midway through the blastula period, 4 hpf) and occurs progressively until 50% epiboly[the onset of gastrulation, 5.3 hpf; see Kimmel et al.(Kimmel et al., 1995) for embryonic stages]. Notably, this cell behaviour predominantly occurs in a region of the dorsal blastoderm margin that is homologous to the organizer epithelium of Xenopus (Cooper and Virta, 2007; Shih and Keller,1992). This suggests that DFCs are formed through a previously unrecognised process of organizer epithelium ingression in zebrafish. Interestingly, this organizer epithelial cell ingression represents the earliest morphogenetic cell behaviour specifically linked to the ontogeny of the organizer region in zebrafish. Furthermore, it adds to the notion that the organizer region can be subdivided into domains of distinct morphogenetic behaviours that generate distinct organ rudiments such as KV, prechordal plate and notochord (Cooper, M. S., Virta, V. C.(2007); Montero et al., 2003). Together, these findings provide novel insights into cell lineage decisions and morphogenesis of the organizer region at the onset of zebrafish gastrulation.

Previous analysis has suggested that KV formation involves a mesenchymal-to-epithelial transition of the DFC cluster(Amack et al., 2007). Here we show that vesicle formation involves an early phase in which epithelial surface cells convert into deep DFCs with mesenchymal behaviour, followed by a phase of re-epithelialisation that transforms the DFC cluster into a ciliated epithelial vesicle. Intriguingly, a subset of polarised DFCs persistently retains epithelial properties throughout the organogenetic process. These DFCs initially form apical contacts with the EVL, which is likely to facilitate their vegetal movement. Subsequently, the polarity of these cells is maintained as they become internalised and form the `seeds' around which the remaining mesenchymal DFCs assemble into polarised rosettes. This observation suggests that epithelial polarity is not initiated de novo within the DFC cluster, but is instead derived from the apicobasal polarity of the surface epithelium. Thus, the initial apicobasal polarity of an embryonic surface tissue provides the polarising cue that underlies the formation of a polarised interior organ. These findings provide important new insights into embryological mechanisms underlying the initiation and propagation of epithelial polarity.

Our finding that KV derives from an embryonic surface epithelium offers new perspectives concerning the evolutionary conservation of laterality organ formation in vertebrates. Similar to KV in zebrafish, the amphibian laterality organ (the gastrocoel roof plate) is a transient ciliated epithelium that forms through internalisation of surface epithelium of the dorsal organizer region (Schweickert et al.,2007; Shook et al.,2004). Moreover, the precursors of the mammalian organ of laterality (the posterior notochordal plate) also appear to be derived from the epithelial epiblast (Blum et al.,2007; Hirokawa et al.,2006; Kinder et al.,2001). However, whereas in Xenopus, where organizer surface epithelial cells internalise through involution, in zebrafish they undergo ingression, suggesting that the cytomechanics of organizer surface epithelium internalisation has evolved differently along phylogenetic lineages(reviewed by Cooper and Virta,2007).

Our evidence for a role of Nodal signalling during organizer epithelium internalisation and DFC specification additionally suggests that the conservation of vertebrate laterality organ formation also extends to the molecular level. In Xenopus, specification of the dorsal organizer epithelium is dependent on Nodal signals produced by the vegetal endoderm(reviewed by Shook et al.,2004). Moreover, specification of the mammalian anterior head process precursors, a group of cells that contributes to the notochordal plate, also requires high levels of Nodal signalling(Vincent et al., 2003; Yamanaka et al., 2007). Together, this strongly points to a conserved role for Nodal signalling during internalisation of outer epithelia, which is likely to be the initial step in vertebrate laterality organ formation. Further studies analysing vertebrate species other than zebrafish will be needed to fully elucidate the conserved cellular and molecular mechanisms by which vertebrate embryos acquired a cilia-dependent, Nodal flow-based strategy to control the handedness of their body plan.

We are particularly grateful to J. Stühmer, M. Krieg, D. Müller,S. Preibisch and P. Tomancak for contributing data to earlier versions of this manuscript; to P. Stockinger for suggesting the use of Tg(sox17:GFP)fish; to J. Peychl and J. Sanderson for microscopy advice; to E. Lehmann, J. Hückmann, G. Junghanns and D. Silva for fish-care; to Y. Arboleda, S. Schneider, R. Pérez and A. Catenaccio for technical help; and to L. Rohde, J. Geiger and M. Tada for valuable advice on earlier versions of the manuscript. This work was supported by grants from the DFG, MPG and EU to C.-P.H., from the PBCT (ACT47), CONICYT/DAAD (2003-4-124), MIDEPLAN (ICM P04-068-F), HHMI (INTNL55005940) and EU (FP6-NEST-PATH EDCBNL) to M.L.C., and from the Proyecto Mecesup UCH0306 and DAAD to P.O.