Specification of the Zebrafish Nervous System by Nonaxial Signals

See allHide authors and affiliations

Science  11 Jul 1997:
Vol. 277, Issue 5323, pp. 254-257
DOI: 10.1126/science.277.5323.254


The organizer of the amphibian gastrula provides the neurectoderm with both neuralizing and posteriorizing (transforming) signals. In zebrafish, transplantations show that a spatially distinct transformer signal emanates from tissues other than the organizer. Cells of the germring (nonaxial mesendoderm) posteriorized forebrain progenitors when grafted nearby, resulting in an ectopic hindbrain-like structure; in contrast, cells of the organizer (axial mesendoderm) caused no posterior transformation. Local application of basic fibroblast growth factor, a candidate transformer in Xenopus, caused malformation but not hindbrain transformation in the forebrain. Thus, the zebrafish gastrula may integrate spatially distinct signals from the organizer and the germring to pattern the neural axis.

The developing vertebrate central nervous system is patterned by inductive interactions (1). The gastrula organizer (referred to as the “dorsal lip” in amphibians, “node” in amniotes, and “shield” in fish) is thought to be the source of patterning information (2). Analyses using amphibian embryos have indicated temporally distinct signals within the organizer (3,4): An activator signal from the anterior axial mesoderm defines the anterior neurectoderm, and a subsequent transformer signal from the chordamesoderm (notochord) repatterns nearby neural tissue into more posterior types. In mouse, chick, and fish embryos, the elimination of the organizer does not abolish anteroposterior (AP) patterning in the neurectoderm; hence, there is a source of pattern information in non-organizer tissues (5, 6).

The neural fate map of zebrafish (7) shows patterning by 6 hours of development, when gastrulation has only advanced to the formation of a thickened blastoderm margin (germring) and an embryonic shield at its dorsal side. Forebrain progenitors are located far from the germring, spanning the dorsal midline (Fig.1A). In contrast, hindbrain progenitors lie close to the germring, lateral to the embryonic shield, with midbrain progenitors in between (Fig. 1A). The early regionalization of anterior (forebrain) and posterior (hindbrain) neural progenitors within the neurectoderm allowed us to investigate signals that may differentially pattern the neuraxis. We hypothesized that proximity to the germring might specify more posterior neural fates. Indeed, labeled presumptive forebrain progenitors (Fig. 1B), transplanted (8) at shield stage to the position of the presumptive hindbrain, adopt the hindbrain fate (Fig. 1C). Moreover, presumptive hindbrain cells are not committed to a specific fate at this stage (9). Thus, the signals that normally instruct or permit cells to adopt the hindbrain fate are still active in vivo at 6 hours.

Figure 1

Transplantation assays identifying signals important for hindbrain specification. (A) Neural fate map of the shield-stage gastrula, created by injecting cells with fluorescent dextran as a lineage tracer [modified from (7)]. (B and C) Proximity to germring tissue may be necessary for hindbrain fate. In (B), the normal fate (determined by homotopic replacement) of cells (white) located at the animal pole is to become telencephalon and retina (ov, otic vesicle). When these cells are transplanted close to the germring (C), they (brown) participate normally in the formation of the hindbrain and express Krox20 (blue) along with other host cells (r3 and r5, rhombomeres 3 and 5). (D to F) Normal descendants of the germring. Fluorescently labeled cells appear white. In (D), a homotopic graft of the central third of the embryonic shield yields labeled cells in the hatching gland (hg), head mesenchyme (hm), notochord (noto), and floorplate (fp), as previously reported (10). In (E), a 90° germring contains mainly somitic progenitors (som), with a minor contribution to endoderm (en) and head mesenchyme. In (F), a 180° germring contains progenitors of posterior trunk and tail somites (som), blood precursors (b), and endoderm. (G to J) Morphological changes in embryos that received mesendoderm grafts to the animal pole. A control embryo (G) is shown in frontal view. In (H), a typical embryo with a 90° germring graft has a left eye (e) that is displaced inward and downward by the ectopic hindbrain-like mass and otic vesicle. Ectopic pigment cells (pg) are present. In a case in which no ectopic otic vesicle developed (I), the graft (white) rests at the center of the anterior neuraxis, and the normal telencephalon is replaced by a tissue mass (bracketed by white arrowheads) resembling the anterior hindbrain, with the characteristic L-shaped bend of the isthmus and large pigment cells (black arrowheads). In an embryo that received a 0° (shield) graft (J), no ectopic otic vesicle or hindbrain-like mass of tissue was seen near the graft. The ventral forebrain (arrowhead) appears to be enlarged. All scale bars, 100 μm.

Because deletion of the shield disrupts notochord but not hindbrain development (6), the signals responsible for hindbrain patterning probably do not come exclusively from the shield. Germring tissue may be a source of such a posteriorizing signal. The shield contributes to axial mesoderm, notochord, and ventral neural tissues (10) (Fig. 1D); the germring gives rise to somitic mesoderm, posterior mesoderm, and endoderm (Fig. 1, E and F) (11). To investigate patterning by nonaxial germring tissue, we transplanted sectors of the shield (0°) and the germring at defined angular distances from the dorsal midline (45°, 90°, 135°, and 180°) to the animal pole, a region fated to become forebrain (Fig. 1, A and B), of shield-stage zebrafish. If the germring were the source of a patterning signal, such grafts should cause presumptive forebrain cells to adopt more posterior fates.

Embryos that received a fluorescently labeled germring graft to the animal pole developed hindbrain-like structures in the forebrain (Fig.1, G to I). Ectopic host-derived structures included otic vesicles (Fig. 1H), a midbrain-hindbrain isthmus, and part of the anterior hindbrain (Fig. 1I). Large pigment cells were found near the graft site (Fig. 1H, arrows). These germring signals have a short-range effect; the contralateral forebrain appeared unaffected in recipients (Fig. 1H; see also Fig. 2C). Transplantations of ectodermal tissue into the animal pole did not cause a hindbrain transformation (12). At present, it is unclear which cell population (somitic mesoderm, endoderm, or both) is responsible for these transformations, because the germring contains a mixed progenitor population.

Figure 2

Molecular changes in the forebrain after germring transplant. (A) Control embryo in which Krox20 (purple) marks rhombomeres 3 and 5 (r3 and r5) in the hindbrain. (B) In the forebrain of an embryo with a 90° germring graft (orange-brown), ectopic Krox20 expression (blue, arrowheads) indicates a transformation to hindbrain. (C) A section of an embryo with a 180° germring graft (e, eye). The graft (white arrowhead) does not overlap with the Krox20-positive domain (black arrowhead). (D) Ectopic Krox20 expression (white arrowhead) can be detected as early as 10 to 11 hours of development, when the host's normal Krox20 expression begins (white arrows). Grafted tissue (yellow arrow) did not express Krox20. (E) Most of the embryos with a shield graft (orange) do not have ectopic Krox20 in the forebrain. In those few that do (not shown), only a few cells are Krox20-positive. (F) Embryos that had received 45° germring grafts, as well as those with 135° and 180° grafts (not shown), show robust ectopic Krox20 expression near the graft in the forebrain. In a population of manipulated embryos, Krox20-positive cells are seen as one band or two bands. An asterisk marks the control embryo. All scale bars, 100 μm.

Expression analysis of Krox20, a zinc-finger transcription factor present in hindbrain rhombomeres 3 and 5 (13) (Fig. 2A), demonstrated that the region close to the germring transplant had adopted a hindbrain-like fate (Fig. 2, B to F, and Table1). Ectopic Krox20 expression was detected as early as 10 to 11 hours of development (the beginning of somitogenesis), when Krox20 normally appears in the host's hindbrain (Fig. 2D); this finding suggested that the regional fate of the forebrain had been altered by the end of gastrulation (10 hours). Ectopic expression appeared as a single patch (Fig. 2, C, D, and F) or as two bands (Fig. 2, B and F). There was no obvious correlation between the origin of the graft and the amount of ectopic Krox20 induced. Ectopic otic vesicles were often found adjacent to the ectopic Krox20 expression, mimicking the normal juxtaposition of otic vesicle and rhombomere 5. Grafted tissue did not contribute to either the ectopic Krox20-expressing cells (Fig. 2C) or the ectopic otic vesicles.

Table 1

Morphological and molecular changes in embryos receiving germring or shield grafts.

View this table:

In contrast, grafting the embryonic shield to the animal pole did not lead to hindbrain-like transformation (Fig. 1J and Table 1). Morphological changes in the forebrain were less drastic than those resulting from other germring grafts; the host's retina was often of near normal size (14) (Fig. 1J). Most of the disruption appeared to result from mechanical obstruction of forebrain morphogenesis, and perhaps from a slight expansion of the forebrain (Fig. 1J, arrowhead). Most embryos with a shield graft did not express any Krox20 in the forebrain (Fig. 2E); a minority had a few Krox20-positive cells near the graft (Table 1) (12).

The germring transplant's ability to change forebrain neural tissue into a section of patterned hindbrain is reminiscent of the “transforming” signal postulated to play a major role in neural patterning in Xenopus. Recent studies in Xenopus(15) have identified basic fibroblast growth factor (bFGF) (4, 16), retinoic acid (RA) (17), and Wnts (18, 19) as candidate transformer signals. Although there is evidence suggesting that RA acts as a posteriorizing signal inXenopus (17), bath (20) or local (21) application of RA does not lead to a transformation of forebrain into hindbrain in zebrafish. Interfering with a number of Wnt molecules and their signaling components can result in substantial caudal defects (22), and there is some evidence that Xwnt3a can synergize with known neural inducers to produce a complete range of AP markers (18). However, no zebrafish Wnt examined to date is expressed in a pattern consistent with a transformer signal (23, 24). The application of bFGF to neuralized animal caps suppresses anterior and enhances posterior neural markers (16), an effect expected of a transformer signal. Furthermore, overexpression of a dominant negative FGF receptor in zebrafish at the early cleavage stages causes a loss of posterior structures (25). These studies suggest that bFGF is a candidate transformer signal in zebrafish.

To determine whether local application of bFGF can mimic the transforming effect of germring tissue transplants, we implanted bFGF beads in the animal pole of shield-stage gastrula (26). The beads caused severe perturbation of the host's forebrain (Fig.3B) but did not cause hindbrain-like transformation (compare Figs. 1H and 3B) or ectopic Krox20 expression (Fig. 3C). Interestingly, the expression of fore-midbrain markerzf-otx2 (27) remained robust (12), which suggested that the regional identity of the forebrain was unchanged. Control beads not coated with bFGF had no effect on brain morphology (Fig. 3A) or gene expression (12). Thus, bFGF, if involved, must act in concert with other factors to generate the germring activity.

Figure 3

Direct test of bFGF as a transformer. (A and B) Morphological changes (left panels, side view; right panels, head-on view) in the forebrain of embryos that received either control or bFGF-coated beads (arrows). Control beads (A) led to no discernible effect on forebrain morphology, whereas bFGF beads (B) caused severe forebrain deformation. In this example, the size of both optic lobes was reduced (right panels), but no hindbrain-like structure or ectopic otic vesicles were observed. (C) Molecular changes in embryos with bFGF beads. Unlike germring grafts (Fig. 2), bFGF does not cause ectopic expression of Krox20 in the forebrain, despite its effect on forebrain morphology (B). Green arrows point to the bead in each embryo. (D) Two physically separated signals acting in the zebrafish gastrula may lead to the observed neuronal fate map. An activator signal originates from the dorsal midline, perhaps before shield formation; this signal acts to induce neural tissue with a broad anterior character (red-graded area). As the germring forms, a transformer signal (blue-graded area) is generated that modulates the axial character of the nearby neural tissue, resulting in distinct forebrain (red), midbrain (not shown), and hindbrain (blue) territories. This transformer signal is strongly present within the germring and is much reduced in the shield. Dorsal ectoderm already exposed to the activator signal can react to the transformer to form the complete nervous system. The activator and transformer signals might be either instructive (acting specifically to promote certain fates) or permissive. The model does not preclude the presence of other signals, for example, those that divide forebrain into telencephalon and diencephalon.

The nonaxial location of transforming activity in the zebrafish suggests a modified two-signal model (Fig. 3D): A general “activating” influence supplied by the organizer (shield) (6) neuralizes the presumptive neurectoderm to produce anterior neural tissue, while a transforming influence abundant in the remainder of the germring (nonaxial mesendoderm) emanates radially to posteriorize the activated (neuralized) tissue. The normal morphogenetic movements during gastrulation (7) displace presumptive forebrain progenitors toward the animal pole (Fig. 3D, white arrow), away from the transforming signal or signals. Hindbrain progenitors, in contrast, remain close to the germring as they converge to the dorsal side (Fig. 3D, black arrows), permitting their transformation to more posterior neural fates. Although this model differs from the classical amphibian activation-transformation model (3, 4) in that the transformer signals reside outside the organizer region, the model is consistent with recent in vitro observations in zebrafish (28). The observation that focally applied bFGF causes head malformation rather than hindbrain transformation in zebrafish gastrula is consistent with the conclusions of a recent transgenic study using Xenopus (29), which indicated that FGFs may not be solely responsible for the neural transforming activity in vivo, in contrast to in vitro results (16). Our study underscores the importance of testing candidate transformer molecules in a relevant spatial and temporal context. In providing evidence for endogenous transformer action in vivo, the model and manipulations presented here can be applied to the dissection of patterning mechanisms in the zebrafish nervous system.

  • * Present address: UCSF/Gallo Center, Building 1, Room 101, San Francisco General Hospital, San Francisco, CA 94110, USA. E-mail: kwoo{at}


View Abstract

Navigate This Article