A dynamic fate map of the forebrain shows how vertebrate eyes form and explains two causes of cyclopia

Forebrain neurulation and common neural-tube defects associated with it,such as cyclopia (Gripp et al.,2000), are best studied dynamically so that normal movements and deviations from them can be directly measured. We have developed new methods to address this class of problem and in particular the early morphogenesis of the eye, using the transparent zebrafish as a model. Fate mapping studies in teleost fish, mouse and chick have shown that a significant anterior-posterior rearrangement of tissue takes place to relocate the hypothalamic anlagen(ventral diencephalon) from a position posterior to the eye field to one anterior and ventral to it (Woo and Fraser, 1995; Varga et al.,1999; Hirose et al.,2004; Inoue et al.,2000; Lawson,1999; Cobos et al.,2001; Dale et al.,1999). This diencephalic rearrangement has been interpreted as splitting the early continuous eye field, with models favouring either a physical separation by a posterior to anterior movement of the future hypothalamus through the plane of the neural plate(Varga et al., 1999; Hirose et al., 2004), or a local induction of medial diencephalic tissue growth to bisect the eye field(Li et al., 1997). Reduced Nodal (Hatta et al., 1991; Hatta et al., 1994; Schier et al., 1996; Rebagliati et al., 1998; Sampath et al., 1998), Wnt11(Heisenberg et al., 1996; Heisenberg and Nüsslein-Volhard,1997; Heisenberg et al.,2000) or Sonic Hedgehog (Chiang et al., 1996) signalling results in cyclopia, a reduced interocular distance or eye fusion, it is presumed by altering forebrain patterning or morphogenesis. We have directly addressed this problem by producing time-lapse confocal movies of wild-type, cyclops(cyc) morphant and homozygous silberblick^tx226(slb) mutant zebrafish embryos. We labelled all cell nuclei with green fluorescent protein (GFP) to visualise and track their movements. In individual embryos, we tracked the paths of hundreds of cells contributing to the forebrain regions of eye, optic stalk, telencephalon, hypothalamus(ventral diencephalon) and dorsal diencephalon to generate high-resolution dynamic fate maps. Movements were followed from mid-gastrula (8 hours post fertilisation, hpf) in the anterior neural plate until the 18-somite stage (15 hpf) when forebrain domains are resolved by morphogenesis and cell fates could be assigned. This enables high-resolution visualisation and quantitative analysis of forebrain morphogenesis for the first time.

Wild-type (AB) embryos were obtained from zebrafish (Danio rerio)lines raised at 28.5°C, maintained as described by Westerfield(Westerfield, 2000). Mutant embryos from homozygous slb carriers were a kind gift of Masazumi Tada (University College London, London, UK). Embryos were staged according to standard indicators (Kimmel et al.,1995).

pCS2-nucGFP2 (a kind gift of Betsy Pownall, University of York, York, UK)was linearised by NotI, and capped mRNA transcripts were made using the MEGAscript SP6 in vitro transcription kit (Ambion). Embryos were injected at the one-cell stage with 80-100 pg of pCS2-nucGPF2 mRNA. cycmorphants were co-injected with 3 ng of cyc morpholino, as previously described (Karlen and Rebagliati,2001).

Suitable embryos were dechorionated and mounted in 0.30% low gelling-temperature agarose (type VII, Sigma Aldrich) containing 0.017%tricaine (Sigma Aldrich) anaesthetic solution(Westerfield, 2000) in custom-built chambers (Concha and Adams,1998) maintained at 28.5°C. Minimal volumes of embryo medium,containing 0.003% tricaine solution, were used to overlay embedded embryos.

Time-lapse imaging was performed using inverted laser scanning confocal microscopes: Zeiss LSM 510 and Leica Microsystems TSC-SP1-MP using long working-distance 20×/0.5 NA water immersion objectives. Image stacks of 50-60 sections of 2.8-3.2 μm spacing were recorded at 2-minute intervals for periods of 20 hours or more. Embryos were morphologically evaluated upon completion of imaging; any displaying abnormal development were excluded.

Manual and automated (G.B.B. and R.J.A., unpublished) cell tracking was performed using custom software written in Interactive Data Language (IDL, ITT Visual Information Solutions). Embryo shape was determined by detecting the embryos surface from its fluorescent signal. A blanket was fitted to this surface and the normals from this used to estimate the centroid of the embryo. Cell locations were expressed relative to this surface, giving a measure of radial depth within the embryo. Measures of the distances between forebrain regions were taken along the curved surface of the embryo, within a plane parallel to the anterior-posterior axis, passing through the centroid. These distances were also converted to relative angles by taking the angular displacement subtended by lines connecting the two locations to the centroid of the embryo. Angular measures were considered less prone to error when comparing measures in living and fixed embryos. Tissue deformation, expressed as contraction and expansion, was measured by calculating local velocity gradients (G.B.B., L.M. and R.J.A., unpublished). For each cell, gradients were calculated with respect to a fixed Cartesian system by looking at the movements of cells and their immediate neighbours. This velocity gradient was used to compute the principal direction and associated principal deformation rates by solving an eigenvalue problem. Cell velocities and positions were corrected for the effects of curvature of the embryo surface in these calculations.

Tracking cells in wild-type animals confirms that hypothalamic precursors originate from a region medial and posterior to the eye field in the anterior neural plate (Fig. 1A)(Varga et al., 1999). The eye field begins as a broad bi-lobed domain bordered anteriorly by precursors of the telencephalon (Fig. 1A, Fig. 2A) and posteriorly by optic stalk. More peripherally there are dorsal diencephalon precursors(Fig. 1A,C) (S.J.E., G.B.B. and R.J.A., unpublished). Optic stalk precursors also originate more anteriorly among eye and telencephalic precursors(Fig. 1A).

Studies in the anterior trunk showed that neurulation in the zebrafish proceeds by folding of the neural plate about the midline to form a transient wedge of cells - the keel - in a mechanism analogous to primary neurulation in higher vertebrates (Papan and Campos-Ortega, 1994; Lowery and Sive, 2004). We show that neurulation within the forebrain is more complex, precluding neurulation through a simple keel or tube. Before neurulation, the most anterior neural plate begins posterior contraction towards a focus at the anterior tip of the hypothalamic anlagen(Fig. 1A, Fig. 3A,D,J,M). The onset of neurulation is detected and defined by a ventral movement of the hypothalamic anlagen, at the focus of contraction, to initiate formation of the anterior limit of the neural keel (Fig. 1B,F, Fig. 3A,D,G,and see Table S1 and Fig. S1 in the supplementary material). This keel subsequently subducts and extends beneath the eye field, without splitting it(Fig. 1C,G), and moves forward to emerge finally anterior to the telencephalic field(Fig. 1D,H, Fig. 3J,M). This posterior-to-anterior rearrangement maps to the dorso-ventral axis of the anterior neural tube (Fig. 3J,M). This movement in the epithelium results in a shear relative to the adjacent lateral tissue. Consequently, posterior eye cells are drawn deep and anteriorwards to begin eye folding(Fig. 1B-D,F-H, Fig. 3G). Coincidently,telencephalon and the anterior-lateral eye field fold towards the midline,which positions them above medial and posterior eye cells(Fig. 1B,F). This creates a neuropore within the dorsal seam of the neural tube near the site of subduction (Fig. 1C,D,G,H),which closes when the lateral edges meet at the midline(Fig. 1I-K,M-O). This folding initiates eye-vesicle evagination by forming an out-pocketing of tissue(Fig. 1I,M). Neurulation of the dorsal forebrain far more closely resembles movements observed in higher vertebrates that involve folding within the neural plate rather than the use of a neural keel.

We find no evidence that the subducting hypothalamus bisects the eye field;in fact, these cells have predominantly converged medially during subduction and remain composed of intermixed precursors of left and right eyes (see Fig. S2A-H in the supplementary material). This is inconsistent with a simple physical displacement by the subducting hypothalamus (ventral diencephalon).

Evagination of the eyes proceeds by a combination of two principal movements. First, the dorsal diencephalon sweeps inwards and forwards,displacing eye and future optic-stalk tissue forwards and then out into the emerging optic vesicles (Fig. 1I-P, Fig. 3P). Second, anterior eye tissue at the midline displaces laterally from above into the eye vesicles as the telencephalon seals the neuropore to occupy the dorsal neural tube (Fig. 1J-K,N-O). This latter movement is predominantly coincident with the segregation of intermingled left and right eye cells into their respective eyes (see Fig. S2I-P in the supplementary material). This description of forebrain morphogenesis is summarised in Fig. 2. 

Given this new model of neural-plate folding in the forebrain, we re-address how neurulation is disrupted in cyclopic mutants. In the cyc mutant, a loss of Ndr2 signalling causes abnormal patterning and morphogenesis of mesoderm and forebrain(Rebagliati et al., 1998; Sampath et al., 1998). After neurulation, a portion of the eye field remains medial as a band of eye tissue spanning the front of the brain between two partial eye cups(Macdonald et al., 1995; Fulwiler et al., 1997). The size and location of the presumptive eye field is normal in cyclopsmutant and morphant embryos; however, hypothalamic tissue is not induced(Fig. 4D, see Fig. S1 in the supplementary material) (Varga et al.,1999). Tracking cells in cyc morphants, we show that the focus of neural-plate contraction (Fig. 3E), and the initiation of neural keel formation, locate within the eye field itself (Fig. 3H,N, Fig. 4E,J,see Table S1 and Fig. S1 in the supplementary material); this is far more anterior to the usual initiation in wild type, which corresponds to the anterior tip of the hypothalamus (Fig. 1B,F). This abnormal focus of contraction causes medial eye-field cells to be brought erroneously anterior and ventral within the neural tube(Fig. 3H,Q, Fig. 4F,L). Morphants additionally show reduced movements of the keel beneath the remaining eye and telencephalic fields: the keel never emerges anterior to them(Fig. 3K,N, Fig. 4F).

A consequence of attenuated subduction movements is that posterior eye tissue does not move forward within the neural plate(Fig. 3K, Fig. 4E,F,P-R). Convergence of telencephalic and anterior eye fields towards the midline is similar to the wild-type (Fig. 4E,F,J,L,P,T),but is not sufficient to resolve the eye-field cells within the keel into separate vesicles (Fig. 4Q,R,V,X). This shows that cyclopia in cyclops results in incorporation of eye tissue into an inappropriate location within the medial neural keel. We speculate that this precludes its resolution into the eye vesicles by any of the morphogenetic mechanisms observed so far in wild-type animals. This represents a dissociation of the mechanisms of patterning and morphogenesis. These defects override potential mechanisms intrinsic to the eye field that usually result in eye separation, such as the motile capacity of eye-field cells (Rembold et al.,2006).

A second mutant, slb, which encodes Wnt11, shows incomplete eye-field separation (Heisenberg et al.,2000). This gene has been associated with reduced movements of convergence and extension, as well as aberrant migration of the mesoderm(Heisenberg and Nüsslein-Volhard,1997; Hammerschmidt et al.,1996; Solnica-Krezel et al.,1996; Ulrich et al.,2003). Effects of Wnt11 on eye morphogenesis could be caused indirectly by disrupting mesodermal induction of neural patterning(Heisenberg and Nüsslein-Volhard,1997; Marlow et al.,1998), or directly from Wnt11 expression in lateral neurectoderm,posterior to the presumptive forebrain(Heisenberg et al., 2000). We investigated eye-field morphogenesis in homozygous slb mutants. Forebrain territories, although patterned normally, are broadened and shortened (Fig. 3F, Fig. 4A, see Fig. S1 in the supplementary material) (Heisenberg et al., 1996), possibly because of defective convergence and extension. The focus of early contraction is less clearly resolved(Fig. 3F). However, the initiation of keel formation is at the anterior boundary of the hypothalamus,as observed in wild type (Fig. 3I,L,O, Fig. 4B,I,see Table S1 and Fig. S1 in the supplementary material). Anteriorward movement of the neural keel, however, is much reduced(Fig. 3L,O, Fig. 4C,K), as in cyclops (Fig. 3K,N),and the hypothalamus never emerges anterior to the telencephalon(Fig. 3L,O, Fig. 4C). Instead, the hypothalamus, optic stalk and posterior eye field are displaced progressively deeper within the keel (Fig. 3I, Fig. 4I,K). Along with a severely reduced convergence of the lateral neural plate, this slows folding of the lateral eye towards the midline(Fig. 4C). Compared with the extensive movement of lateral-posterior eye-field cells fated to optic stalk in wild type (Fig. 1D,H,I-P, Fig. 3P), much reduced convergent and forward movement of the slb eye field results in medial-posterior eye-field cells remaining medial(Fig. 3R, Fig. 4M-O,S,U,W). We postulate that this defect in movement is a major cause of the reduced resolution of bilateral eye fates in this class of mutant, constituting a second, temporally and spatially distinct, defect of forebrain morphogenesis. Analyses of these two mutants reveal two of the intrinsic steps that comprise the complex morphogenetic program of forebrain development.

We thank Bill Harris and members of the Adams group for discussion and critical reading of this manuscript; and many colleagues for providing reagents. We especially thank Masazumi Tada and Nina Buchan for providing silberblick mutant embryos. A Research Studentship from the BBSRC supported S.J.E. A MRC Senior Research Fellowship and Research Grant to R.J.A. supported this work.