Patterning of the Zebrafish Retina by a Wave of Sonic Hedgehog Activity

See allHide authors and affiliations

Science  22 Sep 2000:
Vol. 289, Issue 5487, pp. 2137-2139
DOI: 10.1126/science.289.5487.2137


The Drosophila retina is patterned by a morphogenetic wave driven by the Hedgehog signaling protein. Hedgehog, secreted by the first neurons, induces neuronal differentiation andhedgehog expression in nearby uncommitted cells, thereby propagating the wave. Evidence is presented here that the zebrafish Hedgehog homolog, Sonic Hedgehog, is also expressed in the first retinal neurons, and that Sonic Hedgehog drives a wave of neurogenesis across the retina, strikingly similar to the wave inDrosophila. The conservation of this patterning mechanism is unexpected, given the highly divergent structures of vertebrate and invertebrate eyes, and supports a common evolutionary origin of the animal visual system.

The vertebrate neural retina develops from a layer of pseudostratified epithelium lining the inside of the optic cup, whereas the pigmented retina (RPE) develops from cells on the outside of the cup. The ganglion cell layer (GCL) forms part of the neural retina, and ganglion cells are the first neurons to be born in the retina. Neurogenesis proceeds in a wave from the central to the peripheral retina (1). Sonic Hedgehog (Shh) is expressed in the GCL and RPE and directs proliferation and differentiation of several late arising cell types, such as photoreceptors and glia (2–5). Here we show that at earlier stages, Hedgehog (Hh) signaling drives a wave of shhexpression and neurogenesis across the GCL.

To investigate early functions of Shh in retinal neurogenesis, we constructed a zebrafish strain harboring a green fluorescent protein (GFP) transgene under the control of the Shh promoter (6). Two ShhGFP transformant lines faithfully recapitulate many aspects of shh RNA expression (7, 8). In contrast to the observations of Stenkamp et al. (5), we detect zebrafishshh RNA and ShhGFP not only in the RPE, but also in the GCL (Fig. 1, A and E) (8), which is in agreement with the data from other vertebrates (2–4). ShhGFP and shh RNA expression is activated at 28 to 30 hours in a patch of cells ventral and nasal to the optic disc (Fig. 1B) (8). These cells are the first retinal ganglion cells (RGCs) to differentiate and express the RGC marker Zn5 (Fig. 1B) (9). ShhGFP and shh RNA expression then spreads from this point, together with Zn5 immunoreactivity, and fills the central retina by 52 hours (Fig. 1, A to E) (8). Only a subset of the RGCs express ShhGFP (Fig. 1I).

Figure 1

The shh gene is expressed in a wave in the neural retina. (A to D and F toJ) Confocal micrographs of eyes showing ShhGFP expression (green) and Zn5 staining (red) (32). (E)shh RNA (33). (K and L) Methylene blue–stained sections (33). Anterior is to the left and ventral is down in (B) to (D) and (F) to (J), which are side views of eyes. The broken line demarcates the eye outline. Anterior is up in (A), (E), (K), and (L), which are ventral views of eyes. (A, D, E, H, I, and J) 52 hours; (B and F) 30 hours; (C and G) 40 hours; and (K and L) 76 hours. (A to E, I, and K) Wild type; (F to H, J, and L) syu mutant (13). (I and J) Confocal sections through the GCL, taken at five times the magnification of the other panels. Arrowheads point to the GCL.

To determine whether shh expression might be regulated by Shh itself, we examined ShhGFP expression in sonic you(syu) mutants, in which the zebrafish sonic hedgehoggene is disrupted (10). In syu mutants, ShhGFP expression is initiated in the first RGCs, but then fails to spread further (Fig. 1, F to H). This is very similar to theDrosophila eye, where Hh signaling is required for the spread, but not the induction, of the first Hh-expressing neurons, which instead requires decapentaplegic signaling (11). In contrast, Zn5 immunoreactivity and RGC differentiation do spread in syu mutants, but this spread is retarded (Fig. 1, G and H), and the RGCs are disorganized and reduced in number (Fig. 1, I and J). The reduction of RGCs correlates well with the observation that the optic nerve is thinner in syumutants (7). In addition, the layering ofsyu mutant eyes is not as pronounced as in wild-type eyes (Fig. 1, K and L). At 76 hours, there are many apoptotic cells insyu eyes (Fig. 1L) (7), but elevated cell death is not observed until after 50 hours (7), indicating that cell death is not responsible for the reduced ShhGFP expression, which is already evident well before this (Fig. 1G).

Since Shh is necessary for its own expression, we asked whether it might also be sufficient to induce itself. We injected a shhcDNA under the control of a heat shock–inducible promoter (12) into syu RNA null mutants (13) carrying ShhGFP. DNA injection into zebrafish embryos leads to mosaic expression of the transgene (14). We activated expression by heat shock at 28 hours and examined the effect on ShhGFP at 52 hours. Patches of cells expressing shh RNA were found to induce ShhGFP expression in the GCL (Fig. 2A), indicating that Shh is sufficient to activate its own expression. Consistent with this observation, wild-type cells transplanted into syu eyes are able nonautonomously to induce ShhGFP expression in mutant cells located in the vicinity (Fig. 2B). It is interesting that wild-type clones do not rescue ShhGFP expression if they do not include the region where the wave of neurogenesis starts (Fig. 2C), suggesting that Shh signaling in this area is a prerequisite for the subsequent spread.

Figure 2

Shh induces its own expression in the GCL. (A to C) Confocal micrographs of syu; shhGFP eyes, anterior to the left, ventral to the bottom. The broken line demarcates the eye outline. (A) shh RNA (red) driven by hs-shh (12) induces ShhGFP-expression (green) in the GCL of syu mutants. hs-shh was injected into the RNA null allele of syu (13). (B) Wild-type shhGFP cells labeled with rhodamine (red) and transplanted into syu; shhGFP embryos (34) induce ShhGFP-expression (green) in mutant cells in the vicinity (arrowheads). (C) Wild-type shhGFP cells (red) do not express ShhGFP (green), nor do they rescue ShhGFP-expression in mutant cells if they do not include the point of origin of the neurogenic wave (arrowhead).

These results show that Shh is both necessary and sufficient to control a wave of its own expression that sweeps through the GCL (15). This is strikingly similar to the function of Hh in controlling the morphogenetic furrow of the Drosophila eye (16–18). In contrast to Drosophila, neurogenesis per se is only partially dependent on Shh in the zebrafish retina. As several other Hh genes are known in the zebrafish (19,20), it is possible that one of these might be responsible for the Shh-independent neurogenesis. Consistent with this possibility, we find that tiggywinkle hedgehog (twhh) is expressed in the GCL, and that this expression is detectable, though reduced, insyu eyes (Fig. 3, A and B). To further address this issue, we treated embryos with cyclopamine, which inhibits signaling by both Shh and other Hh family members (21–23). Treatment of embryos with cyclopamine from 26 to 52 hours blocks both the spread of ShhGFP and the spread of neurogenesis (Fig. 3C), indicating that several Hh genes cooperate to drive the wave of neurogenesis in the zebrafish retina.

Figure 3

Contribution of twhh to retinal neurogenesis. (A and B) twhh RNA (33), ventral view of eyes, anterior to the top. (A) Wild type; (B) syu mutant. The arrowheads point to twhhexpression in the GCL. (C and D) Confocal micrographs of cyclopamine-treated eyes, ShhGFP (green), zn5 (red) (32), anterior to the left, ventral to the bottom. The broken line demarcates the eye outline. (C) Treatment with cyclopamine from 26 to 52 hours (13) blocks both the spread of ShhGFP-expression and neurogenesis. (D) Treatment with cyclopamine from 30 to 52 hours blocks the spread of ShhGFP and neurogenesis after it fills the ventral retina.

In Drosophila, Hh is continuously required for furrow progression (17). To determine whether the same might be true in the zebrafish eye, we treated embryos with cyclopamine at later time points. Treatment of embryos from 30 to 52 hours results in eyes in which the spread of ShhGFP and neurogenesis is blocked after it fills a small domain in the ventral anterior retina (Fig. 3D), revealing a continuous requirement for Hh signaling for the neurogenic wave in the zebrafish retina.

In the Drosophila retina, activation of the Ras/MAP-kinase pathway spreads together with the morphogenetic furrow (24), and signaling through the Ras pathway is necessary for retinal neurogenesis, and depends on Hedgehog activity (25). To explore this scenario in the zebrafish retina, we stained eyes with an antibody against the activated form of mitogen-activated protein kinase (dp-ERK) (24). We find that dp-ERK is detectable at 32 hours in the same domain where ShhGFP expression is first activated, and then spreads from this point parallel to ShhGFP expression and neurogenesis (Fig. 4, A to C), as in Drosophila. The spread of dp-ERK occurs in syu eyes, although its domain is smaller (Fig. 4D), whereas it is blocked in embryos treated with cyclopamine from 26 to 52 hours (Fig. 4E).

Figure 4

Activation of ERK spreads in a wave in the neural retina. (A to E) Confocal micrographs of eyes stained for dp-ERK (red) (32), anterior to the left, ventral to the bottom. The broken line demarcates the eye outline. (A) 32 hours; (B) 40 hours; and (C to E) 52 hours. (A to C) Wild type. (D) The wave of dp-ERK activation occurs in syu mutants, although the domain is reduced. (E) The wave of dp-ERK activation is blocked in embryos treated with cyclopamine from 26 to 52 hours (13).

Analysis of the Pax6/Eyeless gene has indicated that the mechanism of eye induction may be conserved across the animal kingdom (26). However, the dramatic variation of eye structure not only between vertebrates and invertebrates, but also within the vertebrate lineage, has suggested that events downstream of eye induction may have evolved independently. Our results now show that the role played by Hh signaling in retinal differentiation is conserved between flies and fish. This suggests that Hh was already used to pattern a primordial eye structure before vertebrate and invertebrate lineages diverged, and thus supports a common evolutionary origin of the animal eye.

  • * To whom correspondence should be addressed. E-mail: carl.neumann{at}


View Abstract

Navigate This Article