The hedgehog (hh) genes code for secreted signaling molecules important for development (Ingham and McMahon, 2001). In zebrafish, Hh signaling from the embryonic midline regulates eye morphogenesis (Ekker et al., 1995; Macdonald et al., 1995); then during retinal differentiation, Hh regulates ganglion cell neurogenesis (Neumann and Neusslein-Volhard, 2000) and Hh from the retinal pigmented epithelium (RPE) regulates photoreceptor differentiation (Stenkamp et al., 2000). The sonic-you (syut4) mutation is a deletion of the sonic hedgehog (shh) gene (Schauerte et al., 1998), one of three hh genes expressed in zebrafish (Currie and Ingham, 1996). Unlike mice lacking shh (Chiang et al., 1996), homozygous mutants are not cyclopic, perhaps because expression of tiggy-winkle hedgehog (twhh) at the midline provides sufficient Hh for dividing the eye fields (Schauerte et al., 1998). However, the eyes of syu- embryos are small (Brand et al., 1996) and have few ganglion cells (Schauerte et al., 1998; Neumann and Neusslein-Volhard, 2000) and differentiated photoreceptors (Stenkamp et al., 2000). We have examined proliferation, cell death, and expression patterns of several genes in the eyes of syu-/- embryos, and find (1) the rate of retinal mitosis is reduced in mutants, but the expression patterns of genes associated with retinal proliferation and neurogenesis (pax6 and ath5) are normal; (2) retinal apoptosis is spatially random and is not initiated until late in retinal development; (3) during the period of retinal differentiation, mutants can be categorized into two distinct retinal phenotypes, i.e., those that become laminated, and those that do not; (4) in nonlaminated mutants, the retina retains neuroepithelial characteristics and differentiated neurons do not appear; (5) in laminated mutants, photoreceptor expression of the zebrafish rx1 gene is reduced, but not that of crx nor neuroD; and initiation of opsin expression in a defined ventral patch is normal, although subsequent spread of expression is limited.
RESULTS AND DISCUSSION
Cell Proliferation and Gene Expression During Retinal Neurogenesis
We examined embryos at 34 hr postfertilization (hpf) by using an antibody (MPM-2) that recognizes cells in M-phase (Escargueil et al., 2000). This time point was selected because it precedes the time of obvious microphthalmia in the mutant (Schauerte et al., 1998; and data not shown), it coincides with a high rate of retinal proliferation, and neurogenesis is in progress (Li et al., 2000a). In each of two experiments, the numbers of M-phase cells/retinal section were lower in mutants than in their wild-type siblings (Fig. 1A–D). A two-way analysis of variance, with clutch and type as “treatment factors” confirmed that, although the results from the two clutches differ, the mutants showed significantly fewer M-phase cells than the wild-type embryos over the two clutches examined. We suspect that minor differences in developmental timing may have contributed to these results, because the rate of proliferation is changing rapidly at this time (Li et al., 2000a). In the second experiment, we also found that the total number of cells in each retinal section was significantly lower in the mutants (Fig. 1E), suggesting reduced proliferation before 34 hpf. The percentage of cells in M-phase was calculated from these data, and was also significantly lower in the mutant eyes (Fig. 1F). Together, these findings indicate a role for Hh signaling in regulating retinal proliferation in the zebrafish, consistent with the in vitro findings for mammals (Jensen and Wallace, 1997; Levine et al., 1997).
Many of the same individual embryos used for analysis of proliferation were also examined for expression patterns of two genes associated with retinal neurogenesis: ath5 (Masai et al., 2000) and pax6 (Hitchcock et al., 1996). Of interest, these patterns were qualitatively normal in the mutant and within the range of normal variation seen in wild-type embryos at 34 hpf (Fig. 1F–I). A regulatory interaction between Hh signaling in the retina and ath5 expression has been predicted (Masai et al., 2000; Kay et al., 2001). Should such an interaction exist, the lack of an alteration in ath5 expression in syu mutants suggests a functional redundancy between shh and twhh. Alternatively, Hh may influence characteristics of retinal neurogenesis independent (or later) than ath5 expression, or a spatiotemporal feature of ath5 expression not evaluated by our experiments.
Most of the remaining analyses examined embryos at 53, 58, and 75 hpf, developmental times when proliferation in wild-type embryos is restricted to the retinal margin and when the retina is composed of differentiating neurons segregated into distinct laminae (see Raymond et al., 1995). In these analyses, 40–50% of the syu-/- embryos examined did not show distinct retinal lamination and, instead, had retinas that appeared to be composed of progenitor cells (see Figs. 2C, 3G, J, and the Experimental Procedures section). These embryos were scored as delayed (Table 1).
Table 1. Embryos Scored as Delayed (Retinas Were Not Laminated) at 53 and 58 hpfa
hpf, hours postfertilization.
53 hpf syu+
53 hpf syu−/−
58 hpf syu+
58 hpf syu−/−
Both wild-type and mutant eyes contained terminal deoxynucleotidyl transferase–mediated dUTP nick end-labeling–positive (TUNEL+) profiles when examined at 34 hpf, near the peak time of cell death in the retina (Cole and Ross, 2001; Fig. 2A,B). Numbers of labeled cells in mutant eyes were not statistically different than those counted in wild-type eyes (Fig. 2G), suggesting that retinal cell survival during this stage of retinal development is not dependent upon Shh signaling. Additionally, this finding suggests that the microphthalmia obvious by 52 hpf (Schauerte et al., 1998) is not likely due to excessive cell death during the period of early neurogenesis. Wild-type eyes showed little or no TUNEL labeling at 53, 58, or 75 hpf (Fig. 2G; see also Cole and Ross, 2001; Biehlmaier et al., 2001). At 53 hpf, cell death was seen in the eyes of six of seven syu-/- embryos (Fig. 2G; no images shown). At 58 hpf, TUNEL labeling was seen in 11 of 14 syu-/- embryos and in some cases was widespread but showed no consistent spatial or laminar pattern (Fig. 2C–G). Therefore, reduced Hh signaling does not selectively induce apoptosis in any specific retinal cell type. The mutants scored as delayed were no more (or less) likely to show extensive cell death (see Fig. 2C). At 75 hpf, extensive TUNEL labeling spanned the entire retina (Fig. 2D,E).
Gene Expression During Retinal Differentiation
Specific transcription factors are involved in photoreceptor development and in regulating expression of photoreceptor-specific genes (pax6, Otteson et al., 2001; neuroD, Yan and Wang, 1998; crx, Chen et al., 1997; rx, Kimura et al., 2000); we examined expression of these transcription factors in the syu-/- mutants to determine whether any of them may mediate the effects of reduced Hh signaling on photoreceptors (Stenkamp et al., 2000; and see below). In 58 hpf syu-/- embryos, there were no qualitative abnormalities in expression of pax6, neuroD, crx, nor of rx2 (Fig. 3A–F; data not shown for rx2). In 5 of 12 mutants (and 0/14 wild-type embryos), these expression patterns were consistent with earlier stages of development (e.g., solid pax6 expression throughout the retina, not shown, or scattered crx expression, Fig. 3G); these individuals were those scored as delayed.
In most of the mutants not scored as delayed, strong crx expression was observed in some cells of the ganglion cell layer (GCL; Fig. 3F), an expression pattern not previously reported for crx (Liu et al., 2001). However, this labeling pattern was also seen in 9 of 14 wild-type embryos (Fig. 3E), and a phenotypical comparison of a syu outcross vs. a (genotypically) wild-type clutch revealed that crx labeling of the GCL is neither associated with the syu allele nor with a specific genetic background (data not shown).
Expression of rx1 was found in progenitor cells and the outer nuclear layer (ONL) and in scattered cells of the inner nuclear layer (INL) of wild-type embryos (Chuang et al., 1999; Fig. 3H). In 7 of 16 mutants, rx1 was expressed in progenitor cells but there was no ONL and these individuals were those scored as delayed by other analyses (Fig. 3J). However, in seven of the remaining nine mutants (those not delayed), a clearly defined ONL was present but these ONL cells did not express rx1 (Fig. 3I). The rx1 gene, therefore, may be a candidate for mediating the effects of reduced Hh signaling on photoreceptor differentiation (Stenkamp et al., 2000, and see below). Alternatively, this expression pattern may be another manifestation of delayed retinal development. We believe the latter alternative is less likely because in mutant embryos not scored as delayed and lacking rx1 expression in the ONL, other photoreceptor markers such as neuroD and crx were expressed in the ONL (Fig. 3D,F).
Expression of Opsins and Immunocytochemical Markers
At 53 hpf, whole-mounted mutant embryos showed patterns of rod and red cone opsin expression similar to those of wild-type embryos and typical for this stage of development (Fig. 4A, B,G; Raymond et al., 1995), suggesting that, at least in some cases, initiation of photoreceptor differentiation in ventral retina was not affected by decreased Hh signaling. However, at 58 hpf, nearly half of the mutant embryos still showed no opsin expression at all (Fig. 4G) and were unlaminated (not shown), consistent with the observation that half of the syu-/- embryos analyzed as cryosections at this developmental time were scored as delayed (Table 1). The remaining mutant embryos showed patterns indistinguishable from those at 53 hpf, whereas wild-type embryo eyes showed a spiral spread of opsin expression throughout the eye (Fig. 4C,D,G; Raymond et al., 1995). At 75 hpf, mutant embryos frequently showed no opsin expression, or expression was still limited to the ventral patch, whereas wild-type eyes showed widespread opsin expression (Fig. 4E,F,G; Raymond et al., 1995). These findings confirm that Hh signaling is important for the propagation of photoreceptor opsin expression (Stenkamp et al., 2000).
Sectioned eyes from 53, 58, and 75 hpf embryos were processed with the retina-specific antibodies zpr-1 (double cones; Larison and Bremiller, 1990), Rho4D2 (rods and green cones; Knight and Raymond, 1990), and RET-1 (ganglion and amacrine cells; Stenkamp et al., 2000). All wild-type and some mutant 53 hpf sections were labeled by both zpr-1 and Rho4D2 in a few cells in ventronasal retina (not shown) and were labeled by RET-1 in the INL and GCL. Wild-type 58 hpf eyes showed extensive labeling by zpr-1 and Rho4D2, whereas in approximately half of the mutants, labeling was absent (Fig. 5G). In the remaining mutants, labeling was confined to the ventronasal retina, even though in nearly all cases a complete ONL was evident (Fig. 5A–D,G), which expressed crx and neuroD (not shown, see Fig. 3D,F). RET-1 labeling in mutants was also reduced or absent (Fig. 5E–G). In 75 hpf wild-type embryos, labeling with zpr-1, Rho4D2, and RET-1 was found throughout the appropriate cell layers, but mutant embryos rarely showed labeling, and in the case of the photoreceptor markers, this labeling was still confined to the ventronasal retina (not shown). A category of none, few, or many (see Experimental Procedures section) was assigned to describe the relative number of cells labeled by each marker in each individual. Complete lack of retina-specific labeling (those scored as none) was primarily or exclusively seen in embryos that were delayed in development (Fig. 5G); however, reduction in retina-specific labeling (those scored as few) was not due to developmental delay. These data indicate a role for Hh signaling in the propagation of rather specific features of retinal differentiation. However, because many retinal transcription factors are expressed normally in the syu-/- embryos (in those which are not delayed), this finding does not support a requirement for Shh in the propagation of neurogenesis per se (but see Neumann and Nusslein-Volhard, 2000).
When examined at 53 or 58 hpf, the eyes of syu-/- embryos showed two distinct phenotypes: approximately half of the syu-/- embryos showed a developmental delay, such that the retina appeared to be composed of progenitor cells rather than differentiating retinal neurons. The remaining syu-/- embryos showed normal retinal lamination but had other unusual attributes. These distinct phenotypes have not been reported previously, possibly because most studies have examined syu-/- embryos at earlier developmental stages, when the phenotype appears more uniform (Brand et al., 1996; Schauerte et al., 1998). Although the appearance of different phenotypes complicates interpretation, it also serves as a useful means of evaluating the role of Hh signaling at different times of development. For example, if all syu-/- mutants were delayed, the role of Hh in photoreceptor differentiation would not have been observed.
In the mutants scored as delayed, the delay may have been related to a decreased rate of proliferation (Fig. 1A–F). The cell proliferation phenotype seen at 34 hpf may foreshadow the appearance of two distinct retinal phenotypes later in development: those that undergo neurogenesis and become laminated, and those that retain a neuroepithelial appearance. The total amount of Hh signal available from twhh expression alone, therefore, may be near a threshold for this effect. If this is the case, Hh signaling may play a role in initiating retinal neurogenesis, perhaps by driving cells through a terminal mitotic step, similar to the situation in the developing Drosophila eye (Heberlein et al., 1993; Ma et al., 1993). It has been noted that an intracellular factor involved in this transition, mitogen-activated protein kinase (dp-ERK; Gabay et al., 1997), also fails to spread through the eyes of syu-/- embryos (Neumann and Nusslein-Volhard, 2000), although the mutants examined in this study initiated dp-ERK expression normally. Of interest, the delayed embryos are unlikely to have any hh expression at all in ganglion cells, because these cells do not differentiate (see Fig. 5G). The delayed phenotype, therefore, may be an indirect consequence of reduced Hh signaling from sources outside the eye, such as axial mesoderm or ventral neural tube. An axial signal of unknown identity has been shown to indirectly participate in initiating retinal neurogenesis, in that zebrafish mutants lacking prechordal plate tissue do not initiate ath5 expression, nor are specific retinal cell types generated (Masai et al., 2000). It is tempting to speculate that this axial signal may be Hh. However, onset of ath5 expression occurs normally in the syu-/- embryos, suggesting that other signals (such as twhh) must be involved, and/or that features of retinal neurogenesis independent of ath5 are regulated by Hh signaling. Furthermore, the axial signal operates indirectly, by means of the optic stalk (Masai et al., 2000), and the optic stalks of syu-/- embryos are known to be unaffected by the mutation (Schauerte et al., 1998). Experiments that more completely knock down Hh signaling in a temporally specific manner are under way to address these issues.
In the mutants not scored as delayed, certain aspects of photoreceptor differentiation (expression of opsins and immunologic markers) were initiated but not propagated. This failure to propagate photoreceptor differentiation is not likely due to photoreceptor cell death, because a clearly defined ONL was present in most cases, which also expressed photoreceptor markers such as neuroD and crx, and because the pattern of cell death appeared to be random. These findings together indicate a role for Hh signaling, probably from the adjacent RPE (Stenkamp et al., 2000), in rather specific features of photoreceptor differentiation. The transcription factor rx1 can be considered a candidate for mediating these effects, because rarely is it expressed in the ONL of mutant eyes and because it is known to participate in regulating expression of photoreceptor-specific genes such as opsin, in vitro (Kimura et al., 2000). However, in adult zebrafish, rx1 expression is restricted to cone photoreceptors (Chuang et al., 1999). If rx1 is similarly restricted to cones in the embryo (which has not been demonstrated), then Hh must act on rod differentiation by other means. An alternative interpretation is that sufficient Hh signaling is required for photoreceptors to fully differentiate and that reduced opsin expression and lack of rx1 in the ONL are both manifestations of incomplete photoreceptor differentiation.
Our studies of cell death, proliferation, and differentiation together offer two explanations for microphthalmia in the syu-/- embryos. Early in retinal development, microphthalmia may be due to a combination of reduced proliferation and failure of half of the mutants to begin the process of retinal neurogenesis, which adds volume to the eye (Li et al., 2000b). Later in retinal development, cell death in the mutant likely contributes to the reduced eye size relative to wild-type eyes. Although cell death in the syu-/- embryo had been observed previously (Schauerte et al., 1998; Neumann and Nusslein-Volhard, 2000), we provide here further documentation that all retinal cell types, including those which have not differentiated, appear to be susceptible to apoptosis. This finding suggests that, in addition to important roles in retinal cell differentiation, adequate Hh signaling is essential for retinal cell survival. Furthermore, because apoptosis (at levels higher than in wild-type eyes) is not detected until after the time that hh expression begins in the RPE (45 hpf; Stenkamp et al., 2000) and well after expression begins in ganglion cells (28 hpf; Neumann and Nusslein-Volhard, 2000), Hh signaling specifically from the RPE or cumulative Hh signaling from several sources may be needed for retinal cell survival. Although syu-/- embryos have circulatory defects (Brand et al., 1996), we believe that cell death in the nervous system is not likely due to ischemia because gas exchange and nutrient distribution by simple diffusion is adequate for zebrafish viability through the first week of life (Stainier and Fishman, 1994; Pelster and Burggren, 1996).
Wild-type zebrafish (Danio rerio) were from our outbred colony and the syut4 strain was from Anand Chandrasekhar (University of Missouri); fish were maintained and bred according to Westerfield (1995). Most embryos were treated with 0.003% phenothiourea at 12 hpf to retard pigmentation (Westerfield, 1995). Homozygous mutants were identified by their curled tails, poor mobility, and circulation defects (Brand et al., 1996); the remaining embryos from a heterozygous cross were considered phenotypically wild-type. Individual embryos were numerically identified and tracked through sectioning and each histologic analysis; only sections passing through the lens in an anterior/posterior orientation were used. Embryos were scored for developmental delay based on absence of retinal lamination (lack of a clearly defined GCL or ONL).
We used the Roche in situ cell death detection kit, with a peroxidase-based signal amplification step. Immunocytochemistry was performed as previously (Stenkamp et al., 2000). Mouse monoclonals zpr-1 (1:200), Rho4D2 (1:200), RET-1 (1:200), and MPM-2 (1:100) were from the University of Oregon zebrafish monoclonal facility, Robert Molday (University of British Columbia), Pamela Raymond (University of Michigan), and Accurate Chemical, respectively. Analysis of MPM-2 and TUNEL labeling on cryosections was done by counting labeled cells per retinal section and averaging results from 1 to 4 sections per individual. These averages were used for subsequent statistical analysis; n-values refer to number of embryos. Retinal cell counts on 34 hpf cryosections were done by using images obtained with Nomarski optics. Analysis of zpr-1, Rho4D2, and RET-1 was done by scoring individual embryos according to whether the relative number of labeled cells could best be described by many (ONL, >15 cells; INL/GCL, >30 cells), few, or none (see Fig. 5G).
Full-length cDNAs (in pBluescript) corresponding to zebrafish shh and twhh, crx, neuroD, pax6 and ath5, rx1 and rx2, and goldfish opsins were the gifts of Steve Ekker (University of Minnesota), Pamela Raymond (University of Michigan), Vladimir Korzh (University of Singapore), Steve Wilson (King's College), Peter Mathers (University of West Virginia), and Koji Nakanishi (Columbia University), respectively. Preparation of riboprobes and in situ hybridization was carried out as previously described (Stenkamp et al., 2000). Retinal coverage by opsin-expressing cells was assessed by using the photoreceptor recruitment stages defined by Raymond et al. (1995).
We thank A. Chandrasekhar for providing syu carriers, and we thank S. Ekker, P. Raymond, V. Korzh, S. Wilson, P. Mathers, and K. Nakanishi for cDNAs, and R. Molday and P. Raymond for providing antibodies. We also thank Pamela Raymond for her critical evaluation of the manuscript, and Kirk Steinhorst for assistance with statistics.
NOTE ADDED IN PROOF
Wang et al.  reported retinal lamination defects following disruption of Hh signaling in mouse ganglion cells. These effects appear to be distinct from our observations that a significant fraction of syu-/- zebrafish retinas fail to become laminated. For example, the zebrafish mutants show no differentiated retinal neurons, while the ganglion cell-Shh-knockout mouse does. Together, these findings provide further evidence for multiple, sequential roles for Hh signaling, from several spatially distinct sources, in regulating retinal development.